Aero-engine assembly precision digital twin model fidelity evaluation method

Abstract

The invention provides a fidelity evaluation method for an aero-engine assembly precision digital twin model. The method comprises the following steps: constructing an assembly precision directed weighted network according to assembly precision characteristics; introducing an assembly precision prediction model used by the digital twin model, and constructing a prediction model fidelity transmission network; identifying fidelity input nodes and the lowest level of the fidelity input nodes of the transmission network, and constructing a fidelity node weight matrix set, a fidelity input matrix and a prediction model weight matrix set; establishing a fidelity transmission and evolution model; and taking the fidelity input matrix, the prediction model weight matrix set and the fidelity node weight matrix set as input, driving a fidelity transmission and evolution model, solving according to an MC method to obtain a fidelity evaluation response, and fitting a fidelity response data set to obtain a fidelity CDF curve and a fidelity PDF curve. According to the method, the fidelity of the aero-engine assembly precision digital twinborn model can be evaluated in real time, and the high-precision and high-consistency digital twinborn model can be selected.

Description

Technical Field

[0001] The present invention belongs to the technical field of aero-engine assembly and digital twin, and specifically relates to a method for evaluating the fidelity of a digital twin model of aero-engine assembly accuracy. Background Art

[0002] The assembly accuracy of aircraft engines has a huge impact on their aerodynamic performance and safety. Given the advantages of digital twin technology such as real-time interaction and dynamic update, digital twin technology has been increasingly used in the assembly accuracy control of aircraft engines. The main control method is to build a digital twin model of aircraft engine assembly accuracy, update the assembly accuracy digital twin model in real time based on the measurement data during the assembly process, and optimize the process parameters on this basis to achieve assembly accuracy control. However, there are many digital twin models of assembly accuracy, and there is currently a lack of evaluation indicators for evaluating the accuracy and consistency of digital twin models of aircraft engine assembly accuracy, which makes it difficult to select models with high accuracy and high consistency, which will seriously hinder the application of digital twins in aircraft engine assembly. Summary of the invention

[0003] The purpose of the present invention is to solve the problem in the prior art that it is difficult to select a high-precision and high-consistency digital twin model of aircraft engine assembly accuracy, and to provide a fidelity evaluation method for a digital twin model of aircraft engine assembly accuracy, which can evaluate the accuracy and consistency of the assembly accuracy digital twin model during the aircraft engine assembly process in real time, and can promote the application of digital twins in aircraft engine assembly.

[0004] To achieve the above purpose, the technical solution provided by the present invention is:

[0005] The present invention provides a method for evaluating the fidelity of a digital twin model of an aircraft engine assembly accuracy, comprising the following steps:

[0006] Step 1, constructing a fidelity transfer network, includes the following sub-steps:

[0007] Step 1.1, extract and abstract the precision features of aircraft engine parts and components into network nodes, determine the direction of each node edge through the assembly relationship, calculate the weight of each node through sensitivity analysis, construct an assembly precision directed weighted network, and obtain the corresponding adjacency matrix;

[0008] Step 1.2, according to the obtained adjacency matrix, introduce the assembly accuracy prediction model used by the assembly accuracy digital twin model, and replace the assembly accuracy node with the fidelity node to construct the assembly accuracy prediction model fidelity transfer network;

[0009] Step 2: construct a fidelity transfer and evolution model based on the assembly accuracy prediction model and the fidelity transfer network. The input of the model is the fidelity input matrix, the fidelity node weight matrix set and the prediction model weight matrix set. The output is the fidelity response data set, which includes the following sub-steps:

[0010] Step 2.1, divide the assembly level according to the distribution of the assembly accuracy prediction model, and construct a set of fidelity node weight matrices by combining the assembly level and the identity matrix;

[0011] Step 2.2, identifying the fidelity input nodes and the lowest level of the fidelity input nodes of the fidelity transfer network at each assembly time, and evaluating the accuracy of the corresponding input data, and combining the evaluation results of the fidelity input nodes and input data to construct a fidelity input matrix;

[0012] Step 2.3, analyzing the assembly accuracy prediction model called by the fidelity transfer network at each moment and evaluating the accuracy of the prediction model, and constructing a prediction model weight matrix set by combining the unit matrix, the prediction model input-output relationship and the prediction model accuracy evaluation result;

[0013] Step 2.4, constructing a digital twin model fidelity transfer and evolution model based on the lowest level of the fidelity input node, the fidelity input matrix, the fidelity node weight matrix set, and the prediction model weight matrix set;

[0014] Step 3: Use the MC sampling method to sample the accuracy evaluation results of the input data at each moment, and drive the fidelity transfer and evolution model to fit the output fidelity response data set to obtain the fidelity CDF curve and fidelity PDF curve, thereby realizing the fidelity evaluation of the digital twin model during the assembly process of the aircraft engine.

[0015] Furthermore, in step 1.2, the fidelity transfer network G at time T1 is first constructed t1 , and then G t1 Extended to all assembly moments T, the assembly accuracy prediction model fidelity transfer network G is obtained t .

[0016] Furthermore, in step 2.2, the fidelity node weight matrix D at time T1 is first constructed. Gω,1 , then the fidelity node weight matrix set is obtained as: D Gω ={D Gω,1 ,D Gω,2 ,…,D Gω,T}.

[0017] Further, step 2.2 includes:

[0018] Step 2.2.1, identifying the fidelity input nodes of the fidelity transfer network at each assembly time, determining the source of the input nodes, and dividing the input data into measured data and theoretical data;

[0019] Step 2.2.2, evaluate the accuracy of the input data, including:

[0020] Step 2.2.2.1, the evaluation method of the measurement data accuracy is expressed as:

[0021] T md =ω me ·T me +ω md ·T md ·δ

[0022] Where, T md Indicates the accuracy of the measured data, T me Indicates the accuracy of the measuring device, ω me represents the measurement equipment accuracy weight, T md represents the impact of data density, ω md represents the data density accuracy weight, δ represents the data density indication parameter;

[0023] Among them, the measurement equipment accuracy T me Calculated by the following formula: In the formula, mv represents the measured value, and te represents the measurement error;

[0024] Data density affects T md Calculated by the following formula: Where md is the data density, ns represents the standard size, a and c are evaluation parameters respectively;

[0025] Step 2.2.2.2, setting the accuracy of the theoretical data to a predetermined value;

[0026] Step 2.2.3, combine the fidelity input nodes at each moment with the input data evaluation results, construct the fidelity input vector at each moment and summarize it into the fidelity input matrix

[0027] Further, step 2.3 includes:

[0028] Step 2.3.1, analyzing the number of prediction models called by the fidelity transfer network at each moment and the input-output relationship of the prediction model;

[0029] Step 2.3.2, the following formula is used by R 2 Methods to evaluate the accuracy of the prediction model:

[0030]

[0031] In the formula, ntest represents the number of test points, y i represents the true measured value, represents the predicted value, represents the mean of the measured values;

[0032] Step 2.3.3, combining the unit matrix, the input-output relationship of the prediction model and the prediction model accuracy evaluation results, construct the prediction model weight matrix set: D pm ={D pm,1 ,D pm,2 ,…,D pm,T}.

[0033] Further, step 2.4 includes:

[0034] Step 2.4.1, express the digital twin model fidelity transfer model as:

[0035]

[0036] In the formula, F 1 represents the fidelity at time T1, represents the fidelity input vector at time T1, D Gω,j and D pm,j Respectively represent the fidelity node weight matrix and prediction model weight matrix to be called;

[0037] Step 2.4.2, express the digital twin model fidelity evolution model as:

[0038]

[0039] In the formula, F i is the fidelity at time i, is the fidelity input vector at time i, L i is the lowest level of input nodes at time i.

[0040] Further, step 3 includes:

[0041] Step 3.1, set the assembly time t = T1, and sample the accuracy evaluation results of the input data at time t. The random variable adopts MC sampling, and the interval variable is equivalent to MC sampling with uniform distribution, and the sampling dimension is q;

[0042] Step 3.2, drive the fidelity transfer and evolution model, calculate the fidelity evaluation response at time t, the dimension is q, and fit the CDF curve and PDF curve of the fidelity at time t;

[0043] Step 3.3, repeat steps 3.1 and 3.2 until the assembly time t = T, and output the fidelity CDF curve and PDF curve at each time.

[0044] Furthermore, in step 1.1, the adjacency matrix is ​​constructed as follows:

[0045]

[0046] In the formula, ω ij Calculated by sensitivity analysis, it represents the weight of the directed edge from node i to j. A value of 0 indicates that there is no direct connection between the corresponding nodes.

[0047] Furthermore, in step 2.2.2.1, ω me and ω md Calculated by analytic hierarchy process.

[0048] The advantages of the present invention are:

[0049] The fidelity evaluation method of the digital twin model of aircraft engine assembly accuracy provided by the present invention first constructs an assembly accuracy directed weighted network according to the accuracy characteristics of assembly parts and components; then introduces the assembly accuracy prediction model used by the digital twin model to construct an assembly accuracy prediction model fidelity transfer network; divides the assembly level according to the prediction model, and constructs a fidelity node weight matrix set; identifies the fidelity input nodes and their lowest levels at each moment, evaluates the accuracy of the input data, and constructs a fidelity input matrix; evaluates the accuracy of the prediction model, and constructs a prediction model weight matrix set; then establishes a fidelity transfer and evolution model; thereafter, the fidelity input matrix, the prediction model weight matrix set, and the fidelity node weight matrix set are used as input to drive the fidelity transfer and evolution model, and the fidelity evaluation response is obtained by solving the MC method, and the fidelity response data set is fitted to obtain a fidelity CDF curve and a PDF curve to achieve fidelity evaluation. Therefore, the present invention can evaluate the fidelity of the digital twin model of aircraft engine assembly accuracy in real time. This method not only helps to discover the limitations and potential improvements of the digital twin model of aircraft engine assembly accuracy, but also can be used as an evaluation index to provide a basis for the selection of twin models, which helps to select high-precision and high-consistency assembly accuracy digital twin models. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The above and / or other features and advantages of the present invention will become more easily understood through the following description with reference to the accompanying drawings, in which:

[0051] Figure 1 It is a flow chart of the method for evaluating the fidelity of the digital twin model of the aircraft engine assembly accuracy of the present invention;

[0052] Figure 2This is the process of constructing the fidelity transfer network at time T1 in the present invention;

[0053] Figure 3 It is the fidelity transfer network of the assembly accuracy prediction model in the present invention;

[0054] Figure 4 It is the fidelity MC evaluation process in the present invention;

[0055] Figure 5 is the casing concentricity directed weighted network adjacency matrix obtained in the example of the present invention;

[0056] Figure 6 It is the fidelity transfer network of the casing concentricity model at time T1 obtained in the example of the present invention;

[0057] Figure 7 It is the evaluation result of the fidelity of the casing concentricity model obtained in the example of the present invention. DETAILED DESCRIPTION

[0058] The present invention will be described in detail below with reference to the accompanying drawings by means of exemplary embodiments of the present invention. It should be noted that the following detailed description of the present invention is only for the purpose of illustration, and is not intended to limit the present invention.

[0059] The present invention provides a method for evaluating the fidelity of an aircraft engine assembly precision digital twin model, which is used to perform real-time evaluation of the accuracy and consistency of the aircraft engine assembly precision digital twin model. It not only helps to discover the limitations and potential improvements of the aircraft engine assembly precision digital twin model, but also can be used as an evaluation indicator to provide a basis for the selection of the twin model.

[0060] First, the overall reference Figure 1 As an exemplary embodiment of the present invention, the fidelity evaluation method of the digital twin model of aircraft engine assembly accuracy generally includes: constructing a fidelity network and a set of fidelity node weight matrices based on sensitivity analysis and combined with the time-varying nature of digital twins; on the basis of the fidelity network, analyzing the fidelity input nodes at each moment, evaluating the accuracy of the input data, and constructing a fidelity input matrix; on the basis of the fidelity network, analyzing the relationship between the assembly accuracy prediction model called at each moment and its input-output relationship, evaluating the accuracy of the prediction model, and constructing a set of prediction model weight matrices; constructing a fidelity transfer and evolution model; and evaluating the fidelity through the MC method.

[0061] Specifically, the method for evaluating the fidelity of the digital twin model of the assembly accuracy of an aero-engine provided by the present invention comprises: step S1, constructing a fidelity transfer network, which specifically comprises the following sub-steps:

[0062] Step S1.1, extract and abstract the precision features of aircraft engine parts and components into network nodes, determine the direction of each node edge through the assembly relationship, calculate the weight of each node through sensitivity analysis, construct an assembly precision directed weighted network, and obtain the corresponding adjacency matrix;

[0063] Step S1.2, based on the obtained adjacency matrix, introduce the assembly accuracy prediction model used by the assembly accuracy digital twin model, replace the assembly accuracy nodes with fidelity nodes, and construct the assembly accuracy prediction model fidelity transfer network.

[0064] The fidelity evaluation method of the present invention further includes: step S2, constructing a fidelity transfer and evolution model based on the assembly accuracy prediction model and the fidelity transfer network, the input of the model is a fidelity input matrix, a fidelity node weight matrix set and a prediction model weight matrix set, and the output is a fidelity response data set. Step S2 specifically includes the following sub-steps:

[0065] Step S2.1, dividing the assembly level according to the distribution of the assembly accuracy prediction model, and constructing a set of fidelity node weight matrices by combining the assembly level and the unit matrix;

[0066] Step S2.2, identifying the fidelity input nodes and the lowest level of the fidelity input nodes of the fidelity transfer network at each assembly time, and evaluating the accuracy of the corresponding input data, and combining the fidelity input nodes and input data evaluation results to construct a fidelity input matrix;

[0067] Step S2.3, analyzing the assembly accuracy prediction model called by the fidelity transfer network at each moment and evaluating the accuracy of the prediction model, and constructing a prediction model weight matrix set by combining the unit matrix, the prediction model input-output relationship and the prediction model accuracy evaluation result;

[0068] Step S2.4, construct the fidelity transfer and evolution model of the digital twin model based on the lowest level of the fidelity input node and the fidelity input matrix, the fidelity node weight matrix set and the prediction model weight matrix set.

[0069] The fidelity evaluation method provided by the present invention also includes: step S3, using the MC sampling method to sample the accuracy evaluation results of the input data at each moment, and driving the fidelity transfer and evolution model to fit the output fidelity response data set to obtain the fidelity CDF curve and the fidelity PDF curve, thereby realizing the evaluation of the fidelity of the digital twin model in the assembly process of the aircraft engine.

[0070] For step S1, Figure 2 As shown in the figure, the connection relationship between the assembly accuracy nodes is determined according to the assembly relationship, and the weights between the nodes are determined through sensitivity analysis. The adjacency matrix of the assembly accuracy in this assembly process is constructed as G:

[0071]

[0072] In the formula, ω ij Calculated by sensitivity analysis, it represents the weight of the directed edge from node i to j. A value of 0 indicates that there is no direct connection between the corresponding nodes.

[0073] Secondly, based on G, the assembly accuracy prediction model used by the digital twin model of aircraft engine assembly accuracy is introduced, and the assembly accuracy node is replaced by the fidelity node to construct the assembly accuracy model fidelity transfer network G at time T1. t1 The sources of fidelity nodes are described by nodes of different shapes. Triangular nodes indicate that the fidelity is determined by measured data, pentagonal nodes indicate that the fidelity is determined by predictions, and parallelogram nodes indicate that the fidelity is determined by theoretical data. Figure 2 In the embodiment of , there are 8 fidelity nodes, the node superscript represents the assembly time, and the subscript represents the assembly precision number, such as: Indicates the fidelity of assembly accuracy a1 at time T1. The fidelity node weight inherits the assembly accuracy node weight. Figure 2 There are three prediction model nodes, namely PM1, PM2 and PM3. The weight of the prediction model node is determined by the accuracy of the prediction model.

[0074] Finally, if Figure 3 As shown, G t1 Extended to all assembly moments, the assembly process in this embodiment is completed through 4 assembly moments, and the assembly accuracy prediction model fidelity transfer network G is constructed t It should be understood that Figure 2 and Figure 3 The assembly accuracy nodes and assembly time etc. used in the fidelity network are only examples for the purpose of convenience of description and are not intended to limit the present invention.

[0075] In step S2, step S2.1 first constructs the fidelity node weight matrix D at time T1 Gω,1 :

[0076]

[0077] Then the fidelity node weight matrix set is obtained as: D Gω ={D Gω,1 ,D Gω,2 ,…,D Gω,T}, each element represents the fidelity node weight matrix at each assembly moment.

[0078] Step S2.2 analyzes the fidelity input nodes at each moment:

[0079] by Figure 3 As an example, we first identify the fidelity input nodes of the fidelity transfer network at each assembly time and the lowest level of the fidelity input nodes, and determine the source of the input nodes. The input data is divided into measured data and theoretical data. The fidelity input nodes at time T1 include and The first three are determined by measured data, and the last two are determined by theoretical tolerance data. According to this method, the fidelity input node is analyzed from the fidelity network at each time.

[0080] Then evaluate the accuracy of the input data:

[0081] The evaluation method of the measurement data accuracy is expressed as:

[0082] T md =ω me ·T me +ω md ·T md ·δ

[0083] Where, T md Indicates the accuracy of the measured data, T me Indicates the accuracy of the measuring device, ω me represents the measurement equipment accuracy weight, T md represents the impact of data density, ω md represents the data density accuracy weight, δ represents the data density indication parameter, if the measurement data accuracy is related to the data density, then δ = 1, otherwise, δ = 0. In some embodiments, ω me and ω md It can be calculated by analytic hierarchy process (AHP).

[0084] Among them, the measurement equipment accuracy T me Calculated by the following formula:

[0085]

[0086] Where mv represents the measured value, te represents the measurement error, and te may be expressed in interval form or random distribution form.

[0087] Data density affects T md Calculated by the following formula:

[0088]

[0089] Where md is the data density, ns represents the standard size, a and c are evaluation parameters, preferably, a=0.15, c=1.

[0090] Theoretical data are usually tolerance data, design data, etc., and their accuracy is Ttd Can be set to a predetermined value, such as T td =0.75.

[0091] Then, the fidelity input nodes at each moment are combined with the input data evaluation results to construct the fidelity input vector at each moment and summarize it into the fidelity input matrix Taking the above embodiment with 4 assembly moments as an example, the fidelity input matrix is ​​as follows:

[0092]

[0093] Where each element is the evaluation result of the input data.

[0094] Step S2.3 may include: Step S2.3.1, analyzing the number of prediction models called by the fidelity transfer network at each moment and the input-output relationship of the prediction models. Figure 2 For example, there are 3 prediction models, which are called by 3 different assembly moments. The input-output relationship of each prediction model is analyzed. Taking prediction model PM1 as an example, its input node is and The output node is

[0095] Step S2.3 also includes: Step S2.3.2, the following formula is obtained by R 2 Methods to evaluate the accuracy of the prediction model:

[0096]

[0097] In the formula, ntest represents the number of test points, y i represents the true measured value, represents the predicted value, Represents the mean of the measurements.

[0098] Then, in step S2.3.3, the prediction model weight matrix at each moment is constructed by combining the unit matrix, the prediction model input-output relationship, and the prediction model accuracy evaluation result. Figure 2 Taking time T1 as an example, the prediction model weight matrix D pm,1 as follows:

[0099]

[0100] According to this method, the prediction model weight matrix at each moment is constructed, and the prediction model weight matrix set is constructed as follows: D pm ={D pm,1 ,D pm,2 ,…,D pm,T}, where each element represents the prediction model weight matrix at each assembly time. It should be noted that Figure 2In the embodiment of FIG. 1 , the prediction model is not called at time T4, so D pm,4 It is the identity matrix. By definition, when the prediction model is not called at this moment, its prediction model weight matrix is ​​the identity matrix.

[0101] In step S2.4, the fidelity transfer model of the digital twin model can be first constructed according to the fidelity input matrix, the fidelity node weight matrix set and the prediction model weight matrix set, and then the fidelity evolution model of the digital twin model can be constructed according to the fidelity input matrix, the fidelity node weight matrix set and the prediction model weight matrix set using the lowest level of the fidelity input node. Specifically, it includes the following sub-steps:

[0102] Step S2.4.1, express the digital twin model fidelity transfer model as:

[0103]

[0104] In the formula, F 1 represents the fidelity at time T1, represents the fidelity input vector at time T1, D Gω,j and D pm,j Respectively represent the fidelity node weight matrix and prediction model weight matrix to be called;

[0105] Step S2.4.2, the digital twin model fidelity evolution model is expressed as:

[0106]

[0107] In the formula, F i is the fidelity at time i, is the fidelity input vector at time i, L i is the lowest level of the input nodes at time i, obtained in step S2.2.

[0108] As for step S3, from the input data evaluation process, it can be seen that the evaluation result of the input data may be in the form of probability distribution or interval. Therefore, the transfer problem of random-interval uncertainty is involved in the fidelity evaluation process. The evaluation process can be as follows: Figure 4 As shown, the specific steps are as follows:

[0109] Step S3.1, set the assembly time t=T1, and sample the accuracy evaluation results of the input data at time t, random variables use MC sampling, interval variables are equivalent to uniform distribution MC sampling, and the sampling dimension is q;

[0110] Step S3.2, drive the fidelity transfer and evolution model, calculate the fidelity evaluation response at time t, the dimension is q, and fit the probability distribution function (CDF) curve and probability density function (PDF) curve of the fidelity at time t;

[0111] Step S3.3, repeat the above steps until the assembly time t=T, and output the CDF curve and PDF curve of the fidelity at each time, thereby realizing the evaluation of the model fidelity during the assembly process.

[0112] As described above, the fidelity evaluation method of the digital twin model of aircraft engine assembly accuracy of the present invention first constructs an assembly accuracy directed weighted network according to the accuracy characteristics of assembly parts and components; then introduces the assembly accuracy prediction model used by the digital twin model to construct an assembly accuracy prediction model fidelity transfer network; divides the assembly level according to the prediction model, and constructs a fidelity node weight matrix set; identifies the fidelity input nodes and their lowest levels at each moment, evaluates the accuracy of the input data, and constructs a fidelity input matrix; evaluates the accuracy of the prediction model, and constructs a prediction model weight matrix set; then establishes a fidelity transfer and evolution model; thereafter, uses the fidelity input matrix, the prediction model weight matrix set, and the fidelity node weight matrix set as inputs to drive the fidelity transfer and evolution model, obtains the fidelity evaluation response by solving the MC method, and fits the fidelity response data set to obtain the fidelity CDF curve and PDF curve to achieve fidelity evaluation. Therefore, the present invention can evaluate the fidelity of the digital twin model of aircraft engine assembly accuracy in real time. This method not only helps to discover the limitations and potential improvements of the digital twin model of aircraft engine assembly accuracy, but also can be used as an evaluation index to provide a basis for the selection of twin models, which helps to select high-precision and high-consistency assembly accuracy digital twin models.

[0113] Next, the aero-engine assembly accuracy digital twin model fidelity evaluation method provided by the present invention is further explained with reference to examples.

[0114] This example takes the fidelity of the digital twin model of the concentricity of the aircraft engine core casing as the object. Those skilled in the art can evaluate the fidelity of other assembly accuracy models, such as the rotor blade tip assembly clearance, multi-stage rotor concentricity, etc.

[0115] First, a directed weighted network of casing concentricity is constructed, which has 26 nodes in total, and the adjacency matrix of the directed weighted network of casing concentricity is obtained as follows: Figure 5 shown.

[0116] The concentricity prediction model takes into account the actual contact of the surface, and its accuracy is 0.9649. The fidelity transfer network of the casing concentricity model at time T1 is constructed as follows: Figure 6 shown.

[0117] The core engine casing assembly involves the front casing, compressor casing and rear casing assembly. The assembly can be divided into two moments, among which the measurement data density is 30. The accuracy of the measurement equipment is shown in Table 1, where L represents the theoretical length of the quantity to be measured.

[0118] Table 1 Measurement equipment accuracy

[0119]

[0120] The data measurement at the two moments is shown in Table 2.

[0121] Table 2 Data measurement at each time

[0122]

[0123] According to AHP, select the measurement equipment accuracy weight ω me =0.4, data density accuracy weight ω md =0.6.

[0124] Then, based on the measurement data accuracy evaluation results and the concentricity model fidelity transfer network, the model fidelity of concentricity at two moments is evaluated according to the fidelity evaluation process, and the CDF curve and PDF curve of the concentricity model fidelity at two moments are obtained as shown in Figure 7 As shown, the fidelity of the casing concentricity digital twin model is evaluated in real time. Therefore, this example verifies the effectiveness of the fidelity evaluation method of the present invention.

[0125] Finally, it should be noted that the features mentioned and / or shown in the above description of the exemplary embodiments of the present invention may be combined in the same or similar manner into one or more other embodiments, combined with the features in other embodiments or substituted for the corresponding features in other implementations. The technical solutions obtained by these combinations or substitutions shall also be deemed to be included in the protection scope of the present invention.

Claims

1. A method for evaluating the fidelity of a digital twin model of aircraft engine assembly accuracy, characterized in that: The following steps are involved: Step 1, constructing a fidelity transfer network, includes the following sub-steps: Step 1.1, extract and abstract the precision features of aircraft engine parts and components into network nodes, The direction of each node edge is determined by the assembly relationship, the weight of each node is calculated by sensitivity analysis, and the assembly accuracy directed weighted network is constructed to obtain the corresponding adjacency matrix; Step 1.2, according to the obtained adjacency matrix, introduce the assembly accuracy prediction model used by the assembly accuracy digital twin model, and replace the assembly accuracy node with the fidelity node to construct the assembly accuracy prediction model fidelity transfer network; Step 2: construct a fidelity transfer and evolution model based on the assembly accuracy prediction model and the fidelity transfer network. The input of the model is the fidelity input matrix, the fidelity node weight matrix set and the prediction model weight matrix set. The output is the fidelity response data set, which includes the following sub-steps: Step 2.1, divide the assembly level according to the distribution of the assembly accuracy prediction model, and construct a set of fidelity node weight matrices by combining the assembly level and the identity matrix; Step 2.2, identifying the fidelity input nodes and the lowest level of the fidelity input nodes of the fidelity transfer network at each assembly time, and evaluating the accuracy of the corresponding input data, and combining the evaluation results of the fidelity input nodes and input data to construct a fidelity input matrix; Step 2.3, analyzing the assembly accuracy prediction model called by the fidelity transfer network at each moment and evaluating the accuracy of the prediction model, and constructing a prediction model weight matrix set by combining the unit matrix, the prediction model input-output relationship and the prediction model accuracy evaluation result; Step 2.4, constructing a digital twin model fidelity transfer and evolution model based on the lowest level of the fidelity input node, the fidelity input matrix, the fidelity node weight matrix set, and the prediction model weight matrix set; Step 3: Use the MC sampling method to sample the accuracy evaluation results of the input data at each moment, and drive the fidelity transfer and evolution model to fit the output fidelity response data set to obtain the fidelity CDF curve and fidelity PDF curve, thereby realizing the fidelity evaluation of the digital twin model during the assembly process of the aircraft engine.

2. The method for evaluating the fidelity of the digital twin model of aircraft engine assembly accuracy according to claim 1 is characterized in that: In step 1.2, first construct the fidelity transfer network G at time T1 t1 , and then G t1 Extended to all assembly moments T, the assembly accuracy prediction model fidelity transfer network G is obtained t .

3. The method for evaluating the fidelity of the digital twin model of aircraft engine assembly accuracy according to claim 2 is characterized in that: In step 2.1, first construct the fidelity node weight matrix D at time T1 Gω,1 , then the fidelity node weight matrix set is obtained as: D Gω ={D Gω,1 ,D Gω,2 ,…,D Gω,T }.

4. The method for evaluating the fidelity of the digital twin model of aircraft engine assembly accuracy according to claim 3 is characterized in that: Step 2.2 includes: Step 2.2.1, identifying the fidelity input nodes of the fidelity transfer network at each assembly time, determining the source of the input nodes, and dividing the input data into measured data and theoretical data; Step 2.2.2, evaluate the accuracy of the input data, including: Step 2.2.2.1, the evaluation method of the measurement data accuracy is expressed as: T md =ω me ·T me +oh md ·T md ·d Where, T md Indicates the accuracy of the measured data, T me Indicates the accuracy of the measuring device, ω me represents the measurement equipment accuracy weight, T md represents the impact of data density, ω md represents the data density accuracy weight, δ represents the data density indication parameter; Among them, the measurement equipment accuracy T me Calculated by the following formula: In the formula, mv represents the measured value, and te represents the measurement error; Data density affects T md Calculated by the following formula: Where md is the data density, ns represents the standard size, a and c are evaluation parameters respectively; Step 2.2.2.2, setting the accuracy of the theoretical data to a predetermined value; Step 2.2.3, combine the fidelity input nodes at each moment with the input data evaluation results, construct the fidelity input vector at each moment and summarize it into the fidelity input matrix 5. The method for evaluating the fidelity of the digital twin model of aircraft engine assembly accuracy according to claim 4 is characterized in that: Step 2.3 includes: Step 2.3.1, analyzing the number of prediction models called by the fidelity transfer network at each moment and the input-output relationship of the prediction model; Step 2.3.2, the following formula is used by R 2 Methods to evaluate the accuracy of the prediction model: In the formula, ntest represents the number of test points, y i represents the true measured value, represents the predicted value, represents the mean of the measured values; Step 2.3.3, combining the unit matrix, the input-output relationship of the prediction model and the prediction model accuracy evaluation results, construct the prediction model weight matrix set: D pm ={D pm,1 ,D pm,2 ,…,D pm,T }.

6. The method for evaluating the fidelity of the digital twin model of aircraft engine assembly accuracy according to claim 5 is characterized in that: Step 2.4 includes: Step 2.4.1, express the digital twin model fidelity transfer model as: In the formula, F 1 represents the fidelity at time T1, represents the fidelity input vector at time T1, D Gω,j and D pm,j Respectively represent the fidelity node weight matrix and prediction model weight matrix to be called; Step 2.4.2, express the digital twin model fidelity evolution model as: In the formula, F i is the fidelity at time i, is the fidelity input vector at time i, L i is the lowest level of input nodes at time i.

7. The method for evaluating the fidelity of the digital twin model of aircraft engine assembly accuracy according to claim 6, characterized in that: Step 3 includes: Step 3.1, set the assembly time t = T1, and sample the accuracy evaluation results of the input data at time t. The random variable adopts MC sampling, and the interval variable is equivalent to MC sampling with uniform distribution, and the sampling dimension is q; Step 3.2, drive the fidelity transfer and evolution model, calculate the fidelity evaluation response at time t, the dimension is q, and fit the CDF curve and PDF curve of the fidelity at time t; Step 3.3, repeat steps 3.1 and 3.2 until the assembly time t = T, and output the fidelity CDF curve and PDF curve at each time.

8. The method for evaluating the fidelity of the digital twin model of aircraft engine assembly accuracy according to claim 1 or 2, characterized in that: In step 1.1, the adjacency matrix is ​​constructed as follows: In the formula, ω ij Calculated by sensitivity analysis, it represents the weight of the directed edge from node i to j. A value of 0 indicates that there is no direct connection between the corresponding nodes.

9. The method for evaluating the fidelity of a digital twin model of aircraft engine assembly accuracy according to any one of claims 3 to 7, characterized in that: In step 2.2.2.1, ω me and ω md Calculated by analytic hierarchy process.

Images