Precision evaluation method and apparatus for engine simulation model, and electronic device

Abstract

The present invention relates to the technical field of engine simulation. Disclosed are a precision evaluation method and apparatus for an engine simulation model, and an electronic device, which are used for providing a technical solution that improves the precision of an engine simulation model to better support the design, optimization and application of engines. The method comprises: acquiring target parameter simulation data and target parameter test data of an engine; determining the target parameter simulation data and the target parameter test data to be target parameter steady-state simulation data, target parameter steady-state test data, target parameter transient-state simulation data and target parameter transient-state test data; on the basis of the target parameter steady-state simulation data and the target parameter steady-state test data, determining the steady-state precision of target parameters; using a plurality of prediction algorithms to determine a fitting degree evaluation result and a similarity evaluation result of the target parameters under each prediction algorithm; and using a grey relational analysis method to determine the transient-state precision of the target parameters.

Description

Method, device and electronic equipment for evaluating the accuracy of an engine simulation model

[0001] This application claims priority to a Chinese patent application filed with the Patent Office of China on March 13, 2024, with application number 202410282665.X and application name “A method, device and electronic device for accuracy assessment of an engine simulation model”, the entire contents of which are incorporated herein by reference. Technical Field

[0002] The present invention relates to the technical field of rocket engine simulation, and in particular to an accuracy evaluation method, device and electronic equipment for an engine simulation model. Background Art

[0003] High-precision and high-reliability aerospace engine modeling and simulation plays an extremely important role in any major aerospace rocket launch and operation activities.

[0004] The diverse and massive data types involved in aerospace engine operation, along with their strong time-varying, highly coupled, and nonlinear characteristics, make traditional model identification and parameter identification techniques inadequate for a high-precision understanding. Consequently, the efficiency and accuracy of aerospace engine operational reliability analysis still need to be improved. Modeling methods based on multi-source information fusion technology can achieve high-precision, real-time, and reliable analysis of liquid rocket engines.

[0005] High-fidelity simulation experiments provide a deeper and more systematic understanding of the internal operating principles and working methods of aerospace engines. This allows for early insight into potential failures and the detection of design flaws, significantly improving development efficiency and quality. Furthermore, simulation reduces the need for physical testing, significantly reducing R&D costs and the risk of failure. Simulation technology also enables rapid technological upgrades and iterations for aerospace engines, driving their continued development.

[0006] Accuracy not only affects the reliability and credibility of simulation models but also their application value. Although my country has built models of rocket engines, simulations are still primarily numerical, leaving room for improvement in accuracy. Accuracy not only affects the reliability and credibility of simulation models but also their application value. Summary of the Invention

[0007] The purpose of the present invention is to provide a method, device and electronic equipment for evaluating the accuracy of an engine simulation model, so as to provide a technical solution for improving the accuracy of the engine simulation model and thus better supporting the design, optimization and application of the engine.

[0008] In a first aspect, the present invention provides a method for evaluating the accuracy of an engine simulation model, the method comprising:

[0009] Acquire target parameter simulation data and target parameter test data of the engine.

[0010] By using a first preset method, the target parameter simulation data and the target parameter test data are determined as target parameter steady-state simulation data, target parameter steady-state test data, target parameter transient simulation data and target parameter transient test data.

[0011] According to a second preset manner, the steady-state accuracy of the target parameter is determined based on the steady-state simulation data of the target parameter and the steady-state test data of the target parameter.

[0012] Based on the target parameter transient simulation data and the target parameter transient test data, a plurality of prediction algorithms are used to determine the fit evaluation results of the target parameters and the similarity evaluation results of the target parameters under each of the prediction algorithms.

[0013] Based on the fit evaluation results of the target parameters corresponding to the multiple prediction algorithms and the similarity evaluation results of the target parameters, the transient accuracy of the target parameters is determined using a grey correlation analysis method.

[0014] Using the above technical solution, the engine simulation model accuracy assessment method of the present invention proposes evaluating the accuracy of the engine simulation model based on steady-state accuracy and transient accuracy. This allows the accuracy of the engine simulation model to be evaluated from different dimensions, improving the accuracy of the engine simulation model assessment and thus better supporting engine design, optimization, and application.

[0015] In the prior art, a single metric is typically used to evaluate engine simulation accuracy. However, since this single metric fails to fully consider the engine's multiple performance indicators and requirements, it can lead to inaccurate assessments of certain key performance parameters, thus affecting simulation accuracy. Furthermore, since various engine performance indicators influence each other, choosing the wrong evaluation metric can also cause deviations in simulation accuracy. Furthermore, for certain complex problems, traditional single-metric evaluation methods struggle to meet multi-objective optimization requirements and fail to provide comprehensive, multi-dimensional simulation accuracy assessments. Finally, traditional single-metric evaluation methods also have shortcomings in evaluating aerospace engine performance under different operating conditions, failing to fully reflect engine performance under these conditions, leading to deviations in evaluation results. The present invention, however, uses a combined prediction accuracy evaluation system based on grey correlation analysis to evaluate the accuracy of engine simulation models. Based on this, the present invention determines the fit evaluation results and similarity evaluation results of the target parameters for each prediction algorithm based on multiple prediction algorithms, thereby providing a more comprehensive assessment of engine performance. By combining the fit evaluation results and similarity evaluation results of the target parameters corresponding to multiple prediction algorithms, the limitations of single algorithms can be mitigated and the accuracy of the evaluation can be further improved. Secondly, combined prediction based on grey relational analysis can effectively reveal the correlations and mutual influences between indicators corresponding to different algorithms. Different engine indicators often have certain correlations, and changing one indicator may affect others. By analyzing the correlations between indicators, this method can more accurately assess the contribution and influence of each parameter, thereby improving the accuracy of the assessment. Furthermore, combined prediction methods based on grey relational analysis can flexibly consider the weights and importance of different indicators. In engine design and optimization, different performance indicators may have different importance and weights. This method can reasonably set weights according to actual needs, thereby more accurately assessing the quality and importance of each indicator, thereby better supporting engine design, optimization, and application.

[0016] Furthermore, the first preset method includes a sliding window algorithm.

[0017] The target parameter steady-state simulation data includes multiple target parameter steady-state simulation data segments; the target parameter steady-state test data includes multiple target parameter steady-state test data segments, the target parameter transient simulation data includes multiple target parameter transient simulation data segments, and the target parameter transient test data includes multiple target parameter transient test data segments; wherein the multiple target parameter steady-state simulation data segments, the multiple target parameter transient simulation data segments, the multiple target parameter transient simulation data segments and the multiple target parameter transient test data segments correspond one to one.

[0018] Furthermore, the second preset method includes determining the steady-state accuracy of the target parameter by using the difference between the steady-state simulation data of the target parameter and the steady-state test data of the target parameter.

[0019] Furthermore, the multiple prediction algorithms include the correlation coefficient method, the root mean square error method, the relative deviation method and the Theil inconsistency coefficient analysis method for determining the fit evaluation results of the target parameters, and the dynamic time warping similarity algorithm, the Procrustes similarity algorithm, the power spectral density algorithm and the Fourier similarity algorithm for determining the similarity evaluation results of the target parameters.

[0020] Furthermore, the determining the transient accuracy of the target parameter by using a grey correlation analysis method based on the fit evaluation results of the target parameter corresponding to the multiple prediction algorithms and the similarity evaluation results of the target parameter includes:

[0021] Normalizing the fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm;

[0022] Determining a target evaluation result from the fit evaluation results of the target parameters corresponding to the multiple prediction algorithms and the similarity evaluation results of the target parameters;

[0023] Determining a correlation coefficient between each of the remaining evaluation results and the target evaluation result based on an absolute difference between the target evaluation result and the remaining evaluation results;

[0024] The transient accuracy of the target parameter is determined based on the correlation coefficients between the remaining evaluation results and the target evaluation result.

[0025] Furthermore, the normalizing of the fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm includes:

[0026] Initializing the data sequence that meets the first preset condition among the fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm;

[0027] The fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm are averaged for the data sequences that meet the second preset condition.

[0028] Furthermore, the first preset condition includes: a difference between at least one data in the data sequence and adjacent data is greater than a first value.

[0029] And / or, the second preset condition includes: the difference between the data in the data sequence and the adjacent data is less than a second value.

[0030] The correlation coefficients between the remaining evaluation results and the target evaluation result satisfy:

[0031] in, Δ ik =|P0(k)-P i (k)|, ρ is the discrimination coefficient, and its value range is [0,1], P0(k) represents the target evaluation result; P i (k) represents the other evaluation results; Δ ik Indicates the absolute value of the difference between the target evaluation result and the other evaluation results, Δ min Indicates the minimum absolute value of the difference between the target evaluation result and the rest of the evaluation results, Δ max Indicates the maximum absolute value of the difference between the target evaluation result and the other evaluation results.

[0032] In a second aspect, the present invention further provides an accuracy assessment device for an engine simulation model, the accuracy assessment device for an engine simulation model comprising:

[0033] The acquisition module is used to acquire target parameter simulation data and target parameter test data of the engine.

[0034] The data determination module is used to determine the target parameter simulation data and the target parameter test data into target parameter steady-state simulation data, target parameter steady-state test data, target parameter transient simulation data and target parameter transient test data using a first preset method.

[0035] The steady-state accuracy determination module is used to determine the steady-state accuracy of the target parameter based on the steady-state simulation data of the target parameter and the steady-state test data of the target parameter in accordance with a second preset method.

[0036] The transient evaluation determination module is used to determine the fit evaluation result of the target parameter and the similarity evaluation result of the target parameter under each prediction algorithm using multiple prediction algorithms based on the target parameter transient simulation data and the target parameter transient test data.

[0037] The transient accuracy determination module is used to determine the transient accuracy of the target parameter based on the fit evaluation results of the target parameter corresponding to the multiple prediction algorithms and the similarity evaluation results of the target parameter using a grey correlation analysis method.

[0038] In a third aspect, the present invention also provides an electronic device comprising: one or more processors; and one or more machine-readable media having instructions stored thereon, which, when executed by the one or more processors, enables the accuracy assessment method of the engine simulation model described in the first aspect to be executed.

[0039] The technical effects achieved by the solutions provided in the second and third aspects are the same as the method solutions provided in the first aspect and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0041] FIG1 is a schematic flow chart of the steps of an accuracy evaluation method for an engine simulation model provided by the present invention;

[0042] FIG2 is a schematic structural diagram of an accuracy evaluation device for an engine simulation model provided by the present invention;

[0043] FIG3 is a schematic structural diagram of an electronic device provided by the present invention;

[0044] FIG4 is a schematic structural diagram of a chip provided by the present invention. DETAILED DESCRIPTION

[0045] To facilitate a clear description of the technical solutions of the embodiments of the present invention, the words "first" and "second" are used in the embodiments of the present invention to distinguish between identical or similar items with substantially the same functions and effects. For example, the first threshold and the second threshold are merely used to distinguish between different thresholds and do not limit their order. Those skilled in the art will understand that the words "first" and "second" do not limit the quantity or execution order, and the words "first" and "second" do not necessarily mean different.

[0046] It should be noted that, in the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the present invention should not be construed as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0047] High-precision and high-reliability aerospace engine modeling and simulation plays an extremely important role in any major aerospace rocket launch and operation activities.

[0048] The diverse and massive data types involved in aerospace engine operation, along with their strong time-varying, highly coupled, and nonlinear characteristics, make traditional model identification and parameter identification techniques inadequate for a high-precision understanding. Consequently, the efficiency and accuracy of aerospace engine operational reliability analysis still need to be improved. Modeling methods based on multi-source information fusion technology can achieve high-precision, real-time, and reliable analysis of liquid rocket engines.

[0049] High-fidelity simulation experiments provide a deeper and more systematic understanding of the internal operating principles and working methods of aerospace engines. This allows for early insight into potential failures and the detection of design flaws, significantly improving development efficiency and quality. Furthermore, simulation reduces the need for physical testing, significantly reducing R&D costs and the risk of failure. Simulation technology also enables rapid technological upgrades and iterations for aerospace engines, driving their continued development.

[0050] Accuracy not only affects the reliability and credibility of simulation models but also their application value. Although my country has built models of rocket engines, simulations are still primarily numerical, leaving room for improvement in accuracy. Accuracy not only affects the reliability and credibility of simulation models but also their application value.

[0051] Based on this, the present invention provides an accuracy assessment method for an engine simulation model. Referring to FIG1 , the accuracy assessment method for an engine simulation model includes:

[0052] S101, obtaining target parameter simulation data and target parameter test data of the engine.

[0053] In an embodiment of the present invention, the target parameters of the engine include but are not limited to thrust, fuel flow, oxidizer flow, combustion chamber pressure, and combustion chamber temperature.

[0054] The target parameter test data is data obtained during engine test. The target parameter simulation data is simulation data corresponding to the target parameter test data. The simulation data can be obtained using simulation software.

[0055] S102 , using a first preset method, determining the target parameter simulation data and the target parameter test data as target parameter steady-state simulation data, target parameter steady-state test data, target parameter transient simulation data, and target parameter transient test data.

[0056] The first preset method may be a sliding window algorithm. The basic idea of ​​the sliding window algorithm is to divide the data into several windows of fixed size and analyze or process the data in each window. Specifically, the sliding window algorithm can be divided into the following steps:

[0057] 1) Interference and Noise Removal: Noise and interference are major factors affecting transient and steady-state analysis of engine parameters. Therefore, filtering, denoising, and noise reduction are required for target parameter steady-state simulation data, target parameter steady-state test data, target parameter transient simulation data, and target parameter transient test data to improve signal quality.

[0058] 2) Determine the window size: First, you need to determine the size of each window. The window size is generally fixed and can be selected according to the actual situation.

[0059] 3) Sliding window: Slide the window from the starting position of the data, slide a fixed step size each time, and repeat the data analysis or processing in the window until it slides to the end of the data.

[0060] 4) Window data processing: Analyze or process the data in each window, such as finding the average value, standard deviation, maximum value, minimum value, etc. of the data in the window.

[0061] 5) Assume that there is a data sequence X with a length of n = {X1, X2, ..., X n}, we need to perform sliding window processing on it, where the window size is k and the sliding step is s. Then, we can use the following formula to express the data W in the window at time i i =[X i ,X i+1 ,...,X i+k-1 ]. The data sequence may be a certain data sequence in the target parameter simulation data and the target parameter test data.

[0062] Among them, [X i , X i+1 ,…,X i+k-1 ] indicates that from X i to X i+k-1 The window of k data points. The sliding window process can be expressed by the following formula: i = i + s;

[0063] Assume that the standard deviation of the data in the window Xi is:

[0064] If the standard deviation σ within a window X If the interval is greater than or equal to δ, where δ is a manually set threshold, the interval is considered steady state; otherwise, it is recorded as transient state. Specifically, the start of the first interval that reaches steady state is considered the beginning of the transient interval, and the end of the first transient interval after the steady state interval minus the step size is considered the end of the steady state interval and the beginning of the transient interval.

[0065] In an embodiment of the present invention, based on the above sliding window algorithm, the target parameter simulation data and the target parameter test data can be determined as target parameter steady-state simulation data, target parameter steady-state test data, target parameter transient simulation data and target parameter transient test data.

[0066] Wherein, the target parameter steady-state simulation data includes multiple target parameter steady-state simulation data segments; the target parameter steady-state test data includes multiple target parameter steady-state test data segments, the target parameter transient simulation data includes multiple target parameter transient simulation data segments, and the target parameter transient test data includes multiple target parameter transient test data segments; wherein, the multiple target parameter steady-state simulation data segments, the multiple target parameter transient simulation data segments, the multiple target parameter transient simulation data segments and the multiple target parameter transient test data segments correspond one to one.

[0067] S103 , determining the steady-state accuracy of the target parameter based on the steady-state simulation data of the target parameter and the steady-state test data of the target parameter in accordance with a second preset manner.

[0068] The second preset method includes: determining the steady-state accuracy of the target parameter by using the difference between the steady-state simulation data of the target parameter and the steady-state test data of the target parameter.

[0069] In practice, steady-state evaluations focus on the errors in engine parameters under steady-state operating conditions. Steady-state accuracy is assessed by comparing simulation results with test data and calculating the difference between the simulation and actual results. This evaluation typically involves comparing key engine parameters such as thrust, combustion chamber pressure, and nozzle temperature. Steady-state accuracy is assessed by calculating the deviation between simulation results and test data.

[0070] S104, based on the target parameter transient simulation data and the target parameter transient test data, using multiple prediction algorithms, determine the fit evaluation results of the target parameters and the similarity evaluation results of the target parameters under each of the prediction algorithms.

[0071] In an embodiment of the present invention, the multiple prediction algorithms include: correlation coefficient method (correlation), root mean square error method (rmse-value), relative deviation method (BIAS) and Theil's U method for determining the fit evaluation results of the target parameters, and dynamic time warping similarity algorithm (DTW), Procrustes similarity algorithm (Procrustes), power spectral density algorithm (PSD) and Fourier similarity algorithm (Fourier) for determining the similarity evaluation results of the target parameters.

[0072] The following is a detailed explanation of each algorithm:

[0073] The core concept of the Dynamic Time Warping (DTW) similarity algorithm is dynamic programming. The DTW algorithm uses an alignment path π to describe the alignment relationship between time series based on three principles: boundedness (the start and end points of the two time series must correspond to each other), monotonicity (data points after an already corresponding data point cannot correspond to data points before an already corresponding data point, i.e., no "cross-correspondence"), and continuity (all data points in one time series must have a corresponding data point in the other time series). The alignment path π is composed of n pairs, each consisting of two data points from two different time series. The alignment path π between these pairs can be defined as follows:

[0074] Given a time series X = {X1, X2, ... XR}, Y = {Y1, Y2, ... YC}, then the alignment path between X and Y is: π = {(π1 X ,π1 Y ),(π2 X ,π2 Y )...(π k X ,π k Y )...(π n X ,π n Y )}

[0075] Since the lengths of different time series are not consistent, there may be a large number of different "one-to-many" alignments between them, so the alignment path is not unique. The goal of the DTW algorithm is to find an alignment path that minimizes the sum of the distance values ​​between corresponding data points and use this sum as the distance value between the series. The DTW algorithm distance is defined as follows:

[0076] Given a time series X = {X1, X2, ...Xm}, Y = {Y1, Y2, ...Yn}, let A denote the set of all aligned paths, then the dynamic time planning distance between X and Y is as follows:

[0077] The DTW algorithm uses dynamic programming theory, viewing the entire solution process as a multi-stage decision-making problem. Starting with the first data point of two time series, the algorithm treats each alignment operation between data points as a decision stage, thereby transforming the calculation of the DTW distance value of the original sequence into the calculation of the distance value of the subsequence. The core of the DTW algorithm is to establish a state transition equation to select the optimal alignment method in each decision stage. The state transition equation is constructed as follows:

[0078] Where DTW(Xi,Yj) represents the DTW distance value required to align the first i data points of time series X with the first j data points of time series Y, and dist(Xi,Yj) represents the Euclidean distance or other distance measurement method (such as Manhattan distance, Chebyshev distance, etc.) between the i-th data point of time series X and the j-th data point of time series Y.

[0079] The state transition equation can be used to calculate each subsequence problem. However, since each subsequence problem depends on three other subsequence problems, and some subsequence problems are dependent on more than one subsequence problem, directly recursively calculating the state transition equation will result in repeated calculations of many subsequence problems. Therefore, the DTW algorithm uses a cumulative distance matrix D to store the values ​​of the subsequence problems, thus avoiding repeated calculations.

[0080] The registration result can be obtained by using the cumulative distance matrix: P = [(A0, B0),…, (A i ,B j ),…,(A m ,B n )], let N be the number of value pairs in the registration result, then P can be written as P = [P1, P2, ..., P N ] According to the obtained registration results, the DTW distance can be calculated according to the following formula:

[0081] Where N is the number of value pairs in the registration result.

[0082] After calculating the DTW distance, the evaluation results of the similarity of the target parameters can be calculated according to the following formula:

[0083] Procrustes Similarity Algorithm. In mathematics, the Procrustes problem generally refers to the problem of matching one object (usually a set of points or shapes) with another. Specifically, given two sets of data points or shapes, the Procrustes method uses transformations such as translation, rotation, and scaling (rigid body transformation) to align one object to the other so that they best match. The term Procrustes includes all classical rigid body motions, as well as the possibility of uniform scaling (stretching or shrinking).

[0084] Specifically, the goal of the Procrustes algorithm is to align two sets of data points so that their average error is minimized. Suppose there are two sets of data points X and Y, where X is an n×d matrix representing n d-dimensional data points, and Y is an n×d matrix representing another set of n d-dimensional data points. The Procrustes algorithm will try to find a rotation matrix R, a scaling factor s, and a translation vector t such that X′=s*X*R+t≈Y

[0085] After normalizing X' and Y, we get X'_norm and Y_norm, let A = X'_norm - Y_norm, and then calculate the similarity evaluation result of the target parameter according to the following formula: Pr ocrustes_simi = 1 - A⊙A

[0086] where ⊙ represents the multiplication operation of corresponding elements, also known as element-wise multiplication or Hadamard product.

[0087] The power spectral density method converts time series into the frequency domain and then calculates their power spectral density to determine the similarity between two time series. Specifically, the power spectral density of two time series can be calculated and then compared for similarity. If the power spectral densities of two time series are similar, then the similarity between them is also high.

[0088] The mathematical implementation of power spectral density can be achieved using Fourier transform. Specifically, converting a time series into the frequency domain can be achieved by performing Fourier transform on it. Then, by calculating the power spectral density of this frequency domain representation, the frequency domain characteristics of the time series can be obtained.

[0089] For a time series x(t) of length N, its Fourier transform is as follows:

[0090] Where f is the frequency, X(f) is the frequency domain representation after Fourier transform, and i is the imaginary unit.

[0091] Using the frequency domain representation X(f) obtained by Fourier transform, the power spectral density P(f) can be calculated using the following formula:

[0092] Where |X(f)| is the amplitude represented in the frequency domain and N is the length of the time series.

[0093] Assuming that after the Fourier transform, their power spectral density functions are X and Y respectively, the similarity evaluation result of the target parameters can be calculated by the following formula: PSD_similarity=|<X,Y*>| / (‖X‖×‖Y‖)

[0094] in:

[0095] <X,Y*> Represents the dot product of X and Y, where Y* represents the conjugate complex number of Y; ||X|| represents the norm of X, that is, the length of vector X; ||Y|| represents the norm of Y, that is, the length of vector Y.

[0096] Smaller similarity values ​​indicate that the curves are more similar, and larger values ​​indicate that the curves are more different.

[0097] The Fourier similarity algorithm is a metric used to compare the similarity between two signals or sequences. It is based on the principle of Fourier transform. The Fourier transform converts the signal from the time domain to the frequency domain, decomposing the signal into a combination of sine and cosine functions, revealing the frequency components of the signal.

[0098] When calculating Fourier similarity, we use the result of Fourier transform, which is the spectrum representation of the two sequences. Suppose we have two sequences x and y, and their spectra after Fourier transform are X and Y respectively. Fourier_similarity = |<X,Y*>| / (‖X‖×‖Y‖)

[0099] in:

[0100] <X,Y*> Represents the dot product of X and Y, where Y* represents the conjugate complex number of Y; ||X|| represents the norm of X, that is, the length of vector X; ||Y|| represents the norm of Y, that is, the length of vector Y.

[0101] The correlation coefficient method calculates the correlation coefficient between two time series, which can be used to assess their similarity. The correlation coefficient ranges from -1 to 1, where the closer the value is to 1, the more similar the two time series are.

[0102] In statistics, the correlation coefficient is a measure of the linear correlation between two variables by calculating the covariance. Covariance is a statistic that measures the relationship between two variables. It can indicate whether the changing trends of the two variables are similar. Covariance can be calculated as follows: cov(X,Y)=E[XE[X]]*E[YE[Y]]

[0103] Where E[X] represents the expected value of variable X, and E[Y] represents the expected value of variable Y. If the covariance is positive, it means that X and Y are positively correlated; if the covariance is negative, it means that X and Y are negatively correlated; if the covariance is 0, it means that there is no linear relationship between X and Y.

[0104] However, the range of covariance values ​​is very large, and it depends on the units and ranges of the variables. Therefore, in order to compare the degree of correlation between different variables, the covariance is usually standardized into a correlation coefficient, which can be calculated as follows: r = cov(X, Y) / (std(X) × std(Y))

[0105] Where cov(X,Y) is the covariance between two real random variables X and Y; std(X) and std(Y) represent the standard deviations of variables X and Y, respectively. This formula ensures that the correlation coefficient ranges from -1 to 1. When the correlation coefficient is 1, it indicates that the two variables are completely positively correlated; when the correlation coefficient is -1, it indicates that the two variables are completely negatively correlated; and when the correlation coefficient is 0, it indicates that there is no linear relationship between the two variables.

[0106] Root Mean Squared Error (RMSE) is a commonly used indicator to measure the prediction error of a regression model. Its calculation formula is:

[0107] Where n is the number of samples, S is the data obtained by simulation, and T is the data obtained by experimental testing.

[0108] RMSE represents the standard deviation of the difference between the predicted value and the true value. The smaller its value, the better the model's predictive ability. Generally speaking, the smaller the RMSE value, the higher the model's predictive accuracy.

[0109] Relative Bias is a metric that measures the difference between simulation data and experimental data and can be used to evaluate the accuracy of simulation data. Relative Bias reflects the average degree of deviation between simulation data and experimental data.

[0110] The formula for calculating relative deviation is as follows:

[0111] Where n is the number of samples, S is the data obtained by simulation, and T is the data obtained by experimental testing.

[0112] Where BIAS ranges from -∞ to +∞. If the relative deviation is 0, it means that the predicted value of the simulation data is completely consistent with the true value of the test data. A larger relative deviation indicates a greater difference between the predicted value of the simulation data and the true value of the test data, and the accuracy of the simulation data is lower.

[0113] 4) Theil's U is a statistical measure used to measure the degree of inconsistency between two variables. It is often used to compare the difference between predicted and true values, and can also be used to compare data from different groups or at different time points.

[0114] The calculation of the Theil inconsistency coefficient analysis method is based on the normalization of the mean squared deviation (MSD). The calculation formula of the Theil inconsistency coefficient analysis method is as follows:

[0115] Among them, P i is the simulation result data, A i is the experimental data, N is the number of data, 0≤U≤1. Theil index analysis method reflects the consistency between the simulation results and the experimental data. U=0 means that the simulation results are completely consistent with the experimental data, and U=1 means that the simulation results are the most different from the experimental data.

[0116] S105 , based on the fit evaluation results of the target parameters corresponding to the multiple prediction algorithms and the similarity evaluation results of the target parameters, a grey correlation analysis method is used to determine the transient accuracy of the target parameters.

[0117] Grey correlation analysis is a method for comparing the correlation between data sequences. It can be used for analysis when the data is small or incomplete. Grey correlation is measured based on the relationship between data sequences. The correlation degree is calculated by normalizing the sequence and calculating the difference between each remaining evaluation result and the reference sequence. The calculated fit evaluation result of the target parameter and the similarity evaluation result of the target parameter are combined into an evaluation matrix to determine a target evaluation result. The evaluation matrix is ​​then initialized and the absolute difference between each remaining evaluation result and the target evaluation result is calculated one by one to obtain the maximum difference and the minimum difference between the two levels. The correlation coefficients of the corresponding elements of each remaining evaluation result and the target evaluation result are then calculated. The weighted average of the correlation coefficients of each data in the remaining evaluation results and the corresponding data of the target evaluation result is calculated to reflect the correlation between each remaining evaluation result and the target evaluation result, i.e., the grey correlation.

[0118] In an embodiment of the present invention, determining the transient accuracy of the target parameter using a grey correlation analysis method based on the fit evaluation results of the target parameter corresponding to the multiple prediction algorithms and the similarity evaluation results of the target parameter includes:

[0119] The fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm are normalized.

[0120] Specifically, normalizing the fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm may include:

[0121] Initialization processing is performed on the data sequence that meets the first preset condition among the fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm.

[0122] The first preset condition includes: the difference between at least one data point in the data sequence and its adjacent data point is greater than a first value. The first value is a value that indicates that the data sequence has a significant increasing and decreasing sequence. The specific value can be determined based on actual conditions and is not specifically limited in this embodiment of the present invention.

[0123] Averaging is performed on the data sequences that meet a second preset condition, among the fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm. The second preset condition includes: the difference between data in the data sequence and adjacent data is less than a second value. This second value is used to indicate that the data sequence does not have a clear upward or downward trend. The specific value can be determined based on actual conditions and is not specifically limited in this embodiment of the present invention.

[0124] A target evaluation result is determined from the fit evaluation results of the target parameters and the similarity evaluation results of the target parameters corresponding to the multiple prediction algorithms.

[0125] It should be understood that the target evaluation result may be either the fit evaluation result of the target parameter or the similarity evaluation result of the target parameter, and this embodiment of the present invention does not specifically limit this. For example, it may be the fit evaluation result of the target parameter determined using the root mean square error method, or it may be the similarity evaluation result of the target parameter determined using the Fourier similarity algorithm.

[0126] Based on the absolute difference between the target evaluation result and the remaining evaluation results, a correlation coefficient of each of the remaining evaluation results is determined.

[0127] Specifically, when the target evaluation result is the fit evaluation result of the target parameters determined by the root mean square error method, the other evaluation results are: the fit evaluation result of the target parameters determined by the correlation coefficient method, the fit evaluation result of the target parameters determined by the relative deviation method, the fit evaluation result of the target parameters determined by the Theil index analysis method, the fit evaluation result of the target parameters determined by the dynamic time warping similarity algorithm, the fit evaluation result of the target parameters determined by the Procrustes similarity algorithm, the fit evaluation result of the target parameters determined by the power spectral density algorithm, and the fit evaluation result of the target parameters determined by the Fourier similarity algorithm.

[0128] Based on this, the correlation coefficient of each of the remaining evaluation results includes: the fit evaluation result of the target parameter determined by the correlation coefficient method, the fit evaluation result of the target parameter determined by the relative deviation method, the fit evaluation result of the target parameter determined by the Theil index analysis method, the fit evaluation result of the target parameter determined by the dynamic time warping similarity algorithm, the fit evaluation result of the target parameter determined by the Procrustes similarity algorithm, the fit evaluation result of the target parameter determined by the power spectral density algorithm, and the absolute value of the difference between the fit evaluation result of the target parameter determined by the Fourier similarity algorithm and the fit evaluation result of the target parameter determined by the root mean square error method.

[0129] It should be understood that the fit evaluation results of the target parameters determined by the correlation coefficient method, the fit evaluation results of the target parameters determined by the relative deviation method, the fit evaluation results of the target parameters determined by the Theil inconsistency coefficient analysis method, the fit evaluation results of the target parameters determined by the dynamic time warping similarity algorithm, the fit evaluation results of the target parameters determined by the Procrustes similarity algorithm, the fit evaluation results of the target parameters determined by the power spectral density algorithm, the fit evaluation results of the target parameters determined by the Fourier similarity algorithm, and the fit evaluation results of the target parameters determined by the root mean square error method are all data sequences, and the number of data in multiple data sequences is the same and corresponds one to one.

[0130] Based on the correlation coefficients between the remaining evaluation results and the target evaluation results, the transient accuracy of the target parameter is determined. The correlation coefficients between the remaining evaluation results and the target evaluation results satisfy:

[0131] in, Δ ik =|P0(k)-P i (k)|, ρ is the discrimination coefficient, and its value range is [0,1], P0(k) represents the target evaluation result; P i (k) represents the other evaluation results, Δ ik Indicates the absolute value of the difference between the target evaluation result and the other evaluation results, Δ min Indicates the minimum absolute value of the difference between the target evaluation result and the rest of the evaluation results, Δ max Indicates the maximum absolute value of the difference between the target evaluation result and the other evaluation results.

[0132] As a possible implementation method, it is known that the time series {x t , t = 1, 2, ..., n} is the original time series, and the predicted value at time t using the above single prediction algorithm is x it, there are m single-item prediction models (i=1, 2, ...m). Using m single-item prediction models and different evaluation criteria, s different combined prediction models are constructed:

[0133] Among them, (α i1 ,α i2 ,...,α im ) T is the weight coefficient vector of the i-th combined prediction model. Generally, if the weight coefficient vectors of s combined prediction models are different, the prediction results will not be exactly the same. Furthermore, the corresponding prediction accuracy evaluation index values ​​will generally be different. To compare these s combined prediction models and select the optimal one, it is necessary to establish a prediction accuracy evaluation index system.

[0134] The eight accuracy evaluation indicators calculated based on the fit evaluation results of the target parameters and the similarity evaluation results of the target parameters reflect the effectiveness of the combined prediction model from different perspectives. For each combined prediction model, the above eight accuracy indicators constitute an evaluation indicator vector (Pi1, Pi2, Pi3, Pi4, Pi5, Pi6, Pi7, Pi8), where i = 1, 2, ... s. This vector records the prediction effects of different individual prediction algorithms. Therefore, the evaluation indicator values ​​of s different combined prediction models constitute an evaluation matrix:

[0135] The fit evaluation result of the target parameters or the similarity evaluation result of the target parameters is: P0 = {P0(1), P0(2)...P0(8)}.

[0136] Since the dimensions of each indicator may be inconsistent, it is necessary to normalize the indicator data. For sequences with obvious increasing or decreasing trends, they can be initialized, that is:

[0137] Among them, P(k) represents each data in the sequence, and P(1) represents the first data in the sequence.

[0138] For sequences with no obvious rising and falling trends, averaging can be used, namely:

[0139] Among them, P(k) represents each data in the sequence, Represents the average of all data in the series.

[0140] After data normalization is completed, the absolute difference between each remaining evaluation result and the target evaluation result needs to be calculated one by one, that is: Δ ik =|P0(k)-P i (k)|

[0141] Then determine the maximum difference between the two levels and the minimum difference between the two levels. The calculation method is as follows:

[0142] Minimum difference between two levels:

[0143] Maximum difference between two levels:

[0144] Then calculate the correlation coefficients of the remaining evaluation results and the target evaluation results respectively:

[0145] Where ρ is the resolution coefficient, which ranges from [0, 1]. The smaller the resolution coefficient, the greater the difference between the correlation coefficients and the stronger the discrimination ability. Generally speaking, ρ is generally set to 0.5.

[0146] Calculate the weighted average of the correlation coefficients of the corresponding elements of the remaining evaluation results and the target evaluation results to reflect the correlation relationship between the remaining evaluation results and the target evaluation result sequence, and call it the correlation degree. The calculation formula is as follows:

[0147] According to the size of the grey weighted correlation degree, the correlation order of each other evaluation result is established. The larger the correlation degree, the more important the other evaluation results are to the evaluation criteria.

[0148] In this embodiment of the present invention, a dataset containing a large number of data samples is obtained, taking a set of time series generated in aerospace engine simulation and a time series obtained from a real physical process as an example. The following shows some of these samples:

[0149] Table 1. Some experimental data and simulation data

[0150] Due to different sampling frequencies, this dataset contains approximately 20,000 samples in the time series generated from aerospace engine simulations, and approximately 100,000 samples in the time series recorded from real physical processes. Each sample records data about the aerospace engine.

[0151] Since there are negative time values ​​in the simulation process, we delete these negative time samples when processing the data to ensure the accuracy and consistency of the data.

[0152] Since simulation data usually contains some noise, and the simulation data itself may have large fluctuations or mutations, it is necessary to filter the simulation data. Filtering can smooth the data, remove noise, and make the changing trend of the data more obvious and visible. Common filtering methods include moving average filtering, median filtering, low-pass filtering, high-pass filtering, etc. In an embodiment of the invention, we use a moving average filtering method to process the simulation data. Moving average filtering is a simple and commonly used filtering method, which smoothes the data by calculating the average value of the data in the window. Specifically, the embodiment of the present invention uses a moving average filter with a window size of 21, and obtains the filtered data by applying the filter to the simulation data. The following is the information of the filter:

[0153] Table 2 Filter information

[0154] The embodiment of the present invention uses a sliding window method to divide the time series into multiple steady-state data segments and transient data segments for the simulation data that has been filtered. By calculating the standard deviation of the observations in the window, the starting point and the end point of the steady-state data segment can be determined. The function uses a sliding window method to slide on the time series in units of step size to calculate the standard deviation of the observations in the window. When the standard deviation of the observations in the window is less than the threshold value 0.001 and was not in a steady state before, the current position is marked as the starting point of the steady-state data segment; when the standard deviation of the observations in the window is greater than or equal to the threshold value and was in a steady state before, the current position is marked as the end point of the transient data segment. Finally, the function returns a list of the starting points and end points of all steady-state data segments.

[0155] Using the sliding window method, we divide the time series into steady-state and transient data segments. Within the steady-state data segment, we observe the following data segments: [1150, 1260], [1410, 50290], and [50780, 99985]. These segments indicate that the time series values ​​within these ranges are relatively stable, with no significant changes or mutations. In contrast, within the transient data segment, we observe the following intervals: [1, 1150], [1260, 1410], and [50290, 50780]. These segments indicate that the time series has experienced significant changes or mutations within these ranges, possibly due to external factors or system characteristics. By identifying and separating steady-state and transient intervals, we can better understand the different characteristics and behaviors of time series data. This also corresponds well to the three stages of aerospace engine startup, water hammer, and shutdown.

[0156] Since the sampling intervals of real data and simulated data are inconsistent, it is possible to identify the time range corresponding to the transient data segments in the real time series and then find the simulation results within this time range. Since the lengths of the two series are almost never the same, interpolation of the simulated series is necessary. Common interpolation methods include linear interpolation, polynomial interpolation, and spline interpolation. This paper selects spline interpolation to fill in missing data points in the simulation series. Spline interpolation uses a polynomial function between local data segments to approximate missing data points, thereby smoothly fitting the entire series. By using spline interpolation, missing values ​​in the simulation series can be filled in as accurately as possible while maintaining data continuity and smoothness.

[0157] It is worth noting that since the DTW algorithm itself is applicable to the case where the two time series are of unequal length, there is no need to interpolate the data before calculating the DTW similarity between the two time series.

[0158] After interpolation, two time series of equal length can be obtained, and the simulation accuracy of each transient interval can be evaluated based on various prediction algorithms.

[0159] The evaluation matrix obtained by calculation is as follows:

[0160] After averaging the columns of the evaluation matrix, the dimensionless evaluation matrix is ​​as follows:

[0161] Select a column as the mother sequence of the evaluation matrix. Here, Fourier similarity is used as the mother sequence and calculated one by one

[0162] The absolute difference between each evaluation object and the parent sequence is calculated as follows:

[0163] The maximum difference between the two levels Δmax = 1.247, and the minimum difference between the two levels Δmin = 0.004. The correlation coefficients of the corresponding elements of each comparison sequence and the parent sequence are calculated as follows:

[0164] For each transient interval, the parameter with the highest grey correlation degree is removed and the remaining parameters are comprehensively evaluated.

[0165] All accuracy evaluation parameters are divided into two sequences, m and n, based on their correlation with accuracy. Sequence m contains m evaluation parameters that are positively correlated with accuracy. Higher values ​​of these evaluation parameters correspond to higher accuracy. Examples include the Fourier similarity algorithm and the correlation coefficient method.

[0166] Sequence n: Contains n evaluation parameters that are negatively correlated with accuracy. The lower the value of these evaluation parameters, the higher the corresponding accuracy. Examples include the root mean square error method, relative deviation method, and Theil index analysis method.

[0167] Calculate as follows:

[0168] The calculated comprehensive accuracy for the interval [1,1150] corresponding to the startup phase is A1 = 0.914, for the interval [1260,1410] corresponding to the water hammer phase is A2 = 0.814, and for the interval [50290,50780] corresponding to the shutdown phase is A3 = 0.771. Since A1 > A2 > A3, it can be assumed that the simulation model for the startup phase has the highest comprehensive accuracy, while the simulation model for the shutdown phase has lower comprehensive accuracy. This is likely due to the shutdown time being advanced by several tens of milliseconds in the simulation model, resulting in significant discrepancies between the simulated and actual values. Based on these results, when modifying the model and improving parameters, the simulation model for the shutdown phase should be the first consideration. The low comprehensive accuracy for the shutdown phase may indicate inconsistencies between the model and the actual data. Therefore, adjusting the simulation model for this interval is key to improving overall simulation accuracy. Furthermore, special attention should be paid to the investigation of time domain errors, as they are one of the main causes of low simulation accuracy during the shutdown phase. By thoroughly analyzing the time-domain errors during the shutdown phase, we can identify discrepancies between the model and actual data, enabling targeted model modifications and parameter improvements. By modifying the shutdown simulation model and exploring the time-domain errors, we can improve the accuracy and reliability of the model during the shutdown phase, enabling it to better reflect real-world conditions. These adjustments help enhance the model's performance in the steady-state range and strengthen the performance and reliability of the entire simulation system. Furthermore, improvements to the shutdown phase will positively impact overall simulation accuracy, increasing the model's practical application value.

[0169] By adaptively adjusting the simulation model within this interval, we can optimize the characteristics of the shutdown phase and the temporal error. Temporal errors can be reduced by adjusting the shutdown time simulation method, optimizing model parameters during the shutdown transition period, and strengthening error compensation during the shutdown process.

[0170] In short, the performance of the simulation model in the shutdown phase can be improved through appropriate methods, and the overall reliability and accuracy of the engine simulation model can be improved by improving the comprehensive accuracy in the shutdown phase.

[0171] In a second aspect, referring to FIG. 2 , the present invention provides an accuracy assessment device for an engine simulation model, the accuracy assessment device for an engine simulation model comprising:

[0172] An acquisition module 201 is used to acquire target parameter simulation data and target parameter test data of the engine;

[0173] The data determination module 202 is configured to determine the target parameter simulation data and the target parameter test data into target parameter steady-state simulation data, target parameter steady-state test data, target parameter transient simulation data, and target parameter transient test data using a first preset method;

[0174] a steady-state accuracy determination module 203 for determining the steady-state accuracy of the target parameter based on the steady-state simulation data of the target parameter and the steady-state test data of the target parameter in accordance with a second preset manner;

[0175] A transient evaluation determination module 204 is configured to determine, based on the target parameter transient simulation data and the target parameter transient test data, a fit evaluation result of the target parameter and a similarity evaluation result of the target parameter under each prediction algorithm using multiple prediction algorithms;

[0176] The transient accuracy determination module 205 is configured to determine the transient accuracy of the target parameter by using a grey correlation analysis method based on the fit evaluation results of the target parameter corresponding to the multiple prediction algorithms and the similarity evaluation results of the target parameter.

[0177] Wherein, the first preset method includes a sliding window algorithm;

[0178] The target parameter steady-state simulation data includes multiple target parameter steady-state simulation data segments; the target parameter steady-state test data includes multiple target parameter steady-state test data segments, the target parameter transient simulation data includes multiple target parameter transient simulation data segments, and the target parameter transient test data includes multiple target parameter transient test data segments; wherein the multiple target parameter steady-state simulation data segments, the multiple target parameter transient simulation data segments, the multiple target parameter transient simulation data segments and the multiple target parameter transient test data segments correspond one to one.

[0179] The second preset method includes: using the difference between the target parameter steady-state simulation data and the target parameter steady-state test data to determine the steady-state accuracy of the target parameter.

[0180] Furthermore, the multiple prediction algorithms include the correlation coefficient method, the root mean square error method, the relative deviation method and the Theil index analysis method for determining the fit evaluation results of the target parameters, and the dynamic time warping similarity algorithm, the Procrustes similarity algorithm, the power spectral density algorithm and the Fourier similarity algorithm for determining the similarity evaluation results of the target parameters.

[0181] The transient evaluation determination module 204 includes:

[0182] The normalization processing unit is used to normalize the fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm.

[0183] The target evaluation result determination unit is used to determine a target evaluation result from the fit evaluation results of the target parameters and the similarity evaluation results of the target parameters corresponding to the multiple prediction algorithms.

[0184] The correlation coefficient determining unit is used to determine the correlation coefficient between each of the remaining evaluation results and the target evaluation result based on the absolute difference between the target evaluation result and the remaining evaluation results.

[0185] A transient accuracy determination unit is used to determine the transient accuracy of the target parameter based on the correlation coefficient between the remaining evaluation results and the target evaluation result.

[0186] The normalization processing unit includes:

[0187] The initialization unit is used to initialize the data sequence that meets the first preset condition in the fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm.

[0188] The averaging unit is used to perform averaging processing on the data sequence that meets the second preset condition in the fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm.

[0189] In a third aspect, an embodiment of the present invention further provides an electronic device comprising: one or more processors; and one or more machine-readable media having instructions stored thereon, which, when executed by the one or more processors, enables the accuracy assessment method of the engine simulation model described in the first aspect to be executed.

[0190] The electronic device in the embodiments of the present invention may be a device, or a component, integrated circuit, or chip in a terminal. The device may be a mobile electronic device or a non-mobile electronic device. For example, the mobile electronic device may be a mobile phone, a tablet computer, a laptop computer, a PDA, an in-vehicle electronic device, a wearable device, an ultra-mobile personal computer (UMPC), a netbook, or a personal digital assistant (PDA), etc. The non-mobile electronic device may be a server, a network attached storage (NAS), a personal computer (PC), a television (TV), a teller machine, or a self-service machine, etc., and the embodiments of the present invention do not specifically limit this.

[0191] The electronic device in the embodiment of the present invention may be a device having an operating system. The operating system may be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present invention.

[0192] FIG3 shows a schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present invention. As shown in FIG3 , the electronic device 400 includes a processor 410 .

[0193] As shown in FIG3 , the processor 410 may be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the present invention.

[0194] As shown in Figure 3, the electronic device 400 may further include a communication line 440. The communication line 440 may include a path for transmitting information between the components.

[0195] Optionally, as shown in FIG3 , the electronic device may further include a communication interface 420. There may be one or more communication interfaces 420. The communication interface 420 may be any transceiver or similar device for communicating with other devices or a communication network.

[0196] Optionally, as shown in FIG3 , the electronic device may further include a memory 430. The memory 430 is used to store computer-executable instructions for executing the solution of the present invention, and is controlled by the processor. The processor is used to execute the computer-executable instructions stored in the memory, thereby implementing the method provided by the embodiment of the present invention.

[0197] As shown in FIG3 , the memory 430 may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, an optical disc storage (including a compact disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 430 may exist independently and be connected to the processor 410 via a communication line 440. The memory 430 may also be integrated with the processor 410.

[0198] Optionally, the computer-executable instructions in the embodiment of the present invention may also be referred to as application program codes, which is not specifically limited in the embodiment of the present invention.

[0199] In a specific implementation, as an embodiment, as shown in FIG3 , the processor 410 may include one or more CPUs, such as CPU0 and CPU1 in FIG3 .

[0200] In a specific implementation, as an embodiment, as shown in Figure 3, the terminal device may include multiple processors, such as the first processor 4101 and the second processor 4102 in Figure 3. Each of these processors may be a single-core processor or a multi-core processor.

[0201] FIG4 is a schematic diagram of the structure of a chip provided by an embodiment of the present invention. As shown in FIG4 , the chip 500 includes one or more (including two) processors 410 .

[0202] Optionally, as shown in FIG4 , the chip further includes a communication interface 420 and a memory 430. The memory 430 may include a read-only memory and a random access memory, and provide operating instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory (NVRAM).

[0203] In some embodiments, as shown in FIG. 4 , the memory 430 stores the following elements: execution modules or data structures, or a subset thereof, or an extended set thereof.

[0204] In an embodiment of the present invention, as shown in FIG4 , a corresponding operation is performed by calling an operation instruction stored in a memory (the operation instruction may be stored in an operating system).

[0205] As shown in FIG. 4 , a processor 410 controls processing operations of any terminal device. The processor 410 may also be referred to as a central processing unit (CPU).

[0206] As shown in FIG4 , memory 430 may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory 430 may also include NVRAM. For example, in an application, the memory, communication interface, and storage are coupled together via a bus system. In addition to a data bus, the bus system may also include a power bus, a control bus, and a status signal bus. However, for clarity, in FIG4 , various buses are labeled as bus system 540.

[0207] As shown in FIG4 , the methods disclosed in the above embodiments of the present invention can be applied to or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits in the processor or by software instructions. The above processor may be a general-purpose processor, a digital signal processor (DSP), an ASIC, a field-programmable gate array (FPGA), or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component. The methods, steps, and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in conjunction with the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor or by a combination of hardware and software modules in the decoding processor. The software modules can be located in a storage medium well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. The storage medium is located in a memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0208] On the one hand, a computer-readable storage medium is provided, in which instructions are stored. When the instructions are executed, the functions performed by the terminal device in the above embodiment are implemented.

[0209] On the one hand, a chip is provided, which is applied to a terminal device. The chip includes at least one processor and a communication interface. The communication interface is coupled to the at least one processor, and the processor is used to run instructions to implement the functions performed by the clock tree generation method in the above embodiment.

[0210] In the above embodiments, all or part of the embodiments may be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer programs or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present invention are performed in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, a terminal, a user device, or other programmable device. The computer program or instructions may be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions may be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium may be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; an optical medium, such as a digital video disc (DVD); or a semiconductor medium, such as a solid-state drive (SSD).

[0211] Although the present invention is described herein in conjunction with various embodiments, in the process of implementing the claimed invention, those skilled in the art can understand and implement other variations of the disclosed embodiments by reviewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple situations. A single processor or other unit can implement several functions listed in the claims. Certain measures are recorded in different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

Claims

1. A method for evaluating the accuracy of an engine simulation model, characterized in that: The accuracy evaluation method of the engine simulation model includes: Acquiring target parameter simulation data and target parameter test data of the engine; Using a first preset method, the target parameter simulation data and the target parameter test data are determined as target parameter steady-state simulation data, target parameter steady-state test data, target parameter transient simulation data, and target parameter transient test data; According to a second preset manner, based on the steady-state simulation data of the target parameter and the steady-state test data of the target parameter, determining the steady-state accuracy of the target parameter; Based on the target parameter transient simulation data and the target parameter transient test data, using multiple prediction algorithms, determine the fit evaluation results of the target parameters and the similarity evaluation results of the target parameters under each of the prediction algorithms; Based on the fit evaluation results of the target parameters corresponding to the multiple prediction algorithms and the similarity evaluation results of the target parameters, the transient accuracy of the target parameters is determined using a grey correlation analysis method.

2. The method for evaluating the accuracy of an engine simulation model according to claim 1, wherein: The first preset method includes a sliding window algorithm; The target parameter steady-state simulation data includes multiple target parameter steady-state simulation data segments; the target parameter steady-state test data includes multiple target parameter steady-state test data segments, the target parameter transient simulation data includes multiple target parameter transient simulation data segments, and the target parameter transient test data includes multiple target parameter transient test data segments; wherein the multiple target parameter steady-state simulation data segments, the multiple target parameter transient simulation data segments, the multiple target parameter transient simulation data segments and the multiple target parameter transient test data segments correspond one to one.

3. The method for evaluating the accuracy of an engine simulation model according to claim 1, wherein: The second preset method includes: using the difference between the target parameter steady-state simulation data and the target parameter steady-state test data to determine the steady-state accuracy of the target parameter.

4. The method for evaluating the accuracy of an engine simulation model according to claim 1, wherein: The multiple prediction algorithms include the correlation coefficient method, the root mean square error method, the relative deviation method and the Theil inconsistency coefficient method for determining the fit evaluation results of the target parameters, and the dynamic time warping similarity algorithm, the Procrustes similarity algorithm, the power spectral density algorithm and the Fourier similarity algorithm for determining the similarity evaluation results of the target parameters.

5. The method for evaluating the accuracy of an engine simulation model according to claim 1, wherein: Determining the transient accuracy of the target parameter by using a grey correlation analysis method based on the fit evaluation results of the target parameter corresponding to the multiple prediction algorithms and the similarity evaluation results of the target parameter includes: Normalizing the fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm; Determining a target evaluation result from the fit evaluation results of the target parameters corresponding to the multiple prediction algorithms and the similarity evaluation results of the target parameters; Determining a correlation coefficient between each of the remaining evaluation results and the target evaluation result based on an absolute difference between the target evaluation result and the remaining evaluation results; The transient accuracy of the target parameter is determined based on the correlation coefficients between the remaining evaluation results and the target evaluation result.

6. The method for evaluating the accuracy of an engine simulation model according to claim 5, wherein: The normalizing of the fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm includes: Initializing the data sequence that meets the first preset condition among the fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm; The fit evaluation results of the target parameters corresponding to each prediction algorithm and the similarity evaluation results of the target parameters corresponding to each prediction algorithm are averaged for the data sequences that meet the second preset condition.

7. The method for evaluating the accuracy of an engine simulation model according to claim 6, wherein: The first preset condition includes: the difference between at least one data in the data sequence and adjacent data is greater than a first value; And / or, the second preset condition includes: the difference between the data in the data sequence and the adjacent data is less than a second value.

8. The method for evaluating the accuracy of an engine simulation model according to claim 5, wherein: The correlation coefficients between the remaining evaluation results and the target evaluation result satisfy: in, Δ ik =|P0(k)-P i (k)|, ρ is the discrimination coefficient, and its value range is [0,1], P0(k) represents the target evaluation result; P i (k) represents the other evaluation results, Δ ik Indicates the absolute value of the difference between the target evaluation result and the other evaluation results, Δ min Indicates the minimum absolute value of the difference between the target evaluation result and the rest of the evaluation results, Δ max Indicates the maximum absolute value of the difference between the target evaluation result and the other evaluation results.

9. An accuracy evaluation device for an engine simulation model, characterized in that: The accuracy evaluation device of the engine simulation model includes: An acquisition module, configured to acquire target parameter simulation data and target parameter test data of the engine; a data determination module, configured to determine, by using a first preset method, the target parameter simulation data and the target parameter test data into target parameter steady-state simulation data, target parameter steady-state test data, target parameter transient simulation data, and target parameter transient test data; a steady-state accuracy determination module, configured to determine the steady-state accuracy of the target parameter based on the steady-state simulation data of the target parameter and the steady-state test data of the target parameter in accordance with a second preset manner; a transient evaluation determination module, configured to determine, based on the target parameter transient simulation data and the target parameter transient test data, a fit evaluation result of the target parameter and a similarity evaluation result of the target parameter under each of the prediction algorithms using a plurality of prediction algorithms; The transient accuracy determination module is used to determine the transient accuracy of the target parameter based on the fit evaluation results of the target parameter corresponding to the multiple prediction algorithms and the similarity evaluation results of the target parameter using a grey correlation analysis method.

10. An electronic device, characterized in that: include: one or more processors; and one or more machine-readable media having instructions stored thereon, which, when executed by the one or more processors, enable execution of the accuracy assessment method for an engine simulation model according to any one of claims 1-8.

Images