Digital twin method for aeroengine vibration troubleshooting

Abstract

The application specifically relates to a digital twin method for troubleshooting of an aero-engine vibration, which mainly comprises the following steps: collecting vibration test data sets of the engine, training test data of each engine by adopting an engine vibration digital twin model to obtain a vibration weight model of each engine; judging vibration states of the ground test data of the engine substituted into the vibration weight model to finally find out the engine with vibration exceeding the standard; and comparing extreme values of assembly parameters and rotor balance parameters of the engine with vibration exceeding the standard to determine the position of the problem part. The purpose is to shorten the development and manufacturing cycle of the engine by introducing environmental parameters, state parameters, control parameters and vibration parameters, constructing a vibration digital twin model based on the engine vibration mechanism and obtaining the correlation between the assembly parameters and the whole machine vibration.

Description

Technical Field

[0001] This invention belongs to the field of aero-engine fault diagnosis technology, specifically relating to a digital twin method for troubleshooting vibration in aero-engines. Background Technology

[0002] Assembly, as the final step in the manufacturing process, determines the quality, performance, and vibration level of an engine. However, in actual engineering, the existence of manufacturing and assembly tolerances, as well as the large number of engine parts, makes it difficult to study the relationship between assembly tolerances and vibration. In practice, vibration exceeding tolerances often occurs due to improper assembly, and it is difficult to pinpoint the specific tolerance parameters as the cause of the deviation. This leads to multiple reworks, increasing labor costs and manufacturing cycles, and affecting factory production efficiency.

[0003] Numerous factors contribute to vibration in aero engines, with the most significant being the imbalance of rotor components. Besides the rotor's inherent imbalance, variations in the relative positions of assembled parts can also lead to rotor imbalance. Studies show that improper tolerances, i.e., improper assembly, can introduce vibration sources, thus affecting the overall engine vibration. This can result in excessive vibration levels during factory testing and inspection. Experienced engineers are then required to troubleshoot the vibration faults. Troubleshooting often necessitates disassembling the engine and meticulously checking each assembly point and component that might be causing the excessive vibration. This process is time-consuming and labor-intensive, and it doesn't always guarantee finding the root cause, leading to multiple rework attempts and sometimes ultimately preventing the product from meeting quality requirements.

[0004] In traditional test runs, the assessment of vibration (mainly horizontal and vertical vibrations of the intake casing, intermediate casing, and turbine rear casing) primarily relies on setting limit values ​​to determine if vibration exceeds the standard. When abnormal vibration is detected, experienced engineers troubleshoot based on test data, assembly parameters, and rotor balance parameters. Subsequently, based on parameter analysis results, components are replaced and reassembled, and a second test run is conducted to verify whether the excessive vibration has decreased to a normal level. If the requirements are met, delivery can proceed; otherwise, component replacement, reassembly, and testing must be repeated until the vibration meets the standard. However, the main drawback of this method is its inability to analyze complex problems. For example, it cannot quantitatively analyze engine vibration levels or compare vibration levels between different engines. Furthermore, this method is limited by human factors, requires significant time and cost, and often fails to yield accurate analysis results. Another approach involves using mathematical statistical methods to analyze the relationship between assembly parameters and overall engine vibration. This includes methods such as principal component analysis, Monte Carlo methods, and multivariate statistical methods to directly calculate the correlation between assembly parameters and overall engine vibration, or using backpropagation (BP) neural networks to build a neural network model between assembly parameters and vibration, thereby enabling vibration prediction and fault monitoring. However, this mathematical statistical analysis method requires a large amount of assembly and vibration data to establish a mathematical relationship between assembly parameters and vibration parameters applicable to all engines of the same model. Therefore, it cannot yield good results for specific problems, and the large amount of assembly data required makes it unsuitable for analysis when data is limited, especially in the early stages of product development. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes a digital twin method for troubleshooting vibration in aero-engines by applying artificial intelligence technology. This method introduces environmental parameters, state parameters, control parameters, and vibration parameters, and constructs a vibration digital twin model based on the engine vibration mechanism to obtain the correlation between assembly parameters and overall engine vibration, thereby shortening the development and manufacturing cycle of the entire engine.

[0006] To achieve the above-mentioned objectives, the technical solution adopted by this invention is: intelligent digital analysis for troubleshooting vibration in aero-engines, comprising the following steps:

[0007] S1: Collect the vibration test dataset of the engine. The vibration test dataset is a collection of test data from several engines. Then, filter, analyze and process the data in the vibration test dataset.

[0008] S2: The test data of each engine is trained using the engine vibration digital twin model to obtain the vibration weight model of each engine;

[0009] S3: Calculate the accuracy of each vibration weight model;

[0010] S4: If the accuracy of the vibration weight model is within the threshold range, proceed to the next step; if the accuracy of the vibration weight model exceeds the threshold range, return to step S1 to check, analyze and process the data and retrain the model.

[0011] S5: Substitute the ground test data of all engines into each vibration weight model in turn to carry out vibration digitization test and obtain the vibration digitization test matrix.

[0012] S6: Analyze the vibration digitization test matrix in step S5 according to the vibration state of the engine, and identify the engines with excessive vibration.

[0013] S7: Compare the extreme values ​​of the assembly parameters and rotor balance parameters of the engine with excessive vibration to the corresponding extreme values ​​of the assembly parameters and rotor balance parameters of other engines to determine the location of the problematic part.

[0014] Preferably, the engine vibration digital twin model described in step S2 includes a feature input layer, a coupling network layer, and a vibration mapping network layer; the coupling network layer extracts all parameters from the feature input layer, couples them, and then transmits them to the vibration mapping network layer; the feature input layer includes feature parameters input by each part or component within the engine.

[0015] Preferably, the engine vibration digital twin model described in step S2 is constructed in sequence into three vibration output sub-networks based on the internal physical operating mechanism or gas flow direction of the engine. The vibration output sub-network includes a feature input layer, a coupling network layer, and a vibration mapping network layer. The feature input layer of the subsequent vibration output sub-network contains all the feature parameters in the feature input layer of the preceding vibration output sub-network.

[0016] Preferably, the specific method for identifying engines with excessive vibration in step S6 is as follows: compare the values ​​of several vibration characteristic parameters obtained by each engine in the same vibration weighting model. If the vibration characteristic parameter value is greater than a set value, the engine is determined to be an engine with excessive vibration.

[0017] Preferably, the specific method for identifying engines with excessive vibration in step S6 is as follows:

[0018] S6.1: Obtain the values ​​of several vibration characteristic parameters of each engine in the same vibration weighting model, and calculate the mean value of each vibration parameter;

[0019] S6.2: Sum the mean values ​​of vibration parameters obtained from various vibration weighting models for the same engine to obtain the engine's vibration score;

[0020] S6.3: Determine whether the vibration score of each engine reaches or exceeds the set value. If so, it is considered an engine with excessive vibration.

[0021] Preferably, in step S7, the specific method for determining the location of the problematic part is as follows: the assembly parameters and rotor balance parameters of the engine with excessive vibration that has been found are compared with the assembly parameters and rotor balance parameters of all other engines; if the assembly parameters and rotor balance parameters of the engine with excessive vibration are the largest, then the part corresponding to the parameters is considered to be the problematic part in the engine with excessive vibration.

[0022] Preferably, in step S7, the specific method for determining the location of the problematic part is as follows: using the Pearson correlation coefficient to calculate the correlation between vibration parameters, assembly parameters, and rotor balance parameters, and finding the assembly parameters and rotor balance parameters that lead to high vibration scores, the parts corresponding to these parameters are the problematic parts.

[0023] Preferably, the assembly parameters include the bearing clamping nut torque, bearing inner and outer ring fit clearance, coaxiality, end face runout value, and cylindrical surface runout value; the rotor balance parameters include the rotor balance parameters of rotating components such as the fan, compressor, low-pressure turbine, and high-pressure turbine.

[0024] An electronic device, characterized in that it comprises:

[0025] One or more processors;

[0026] Storage device for storing one or more programs;

[0027] When the one or more programs are executed by the one or more processors, the one or more processors implement the digital twin method for troubleshooting vibrations in aircraft engines as described above.

[0028] A computer-readable medium storing a computer program, characterized in that: when the computer program is executed by a processor, it implements the digital twin method for troubleshooting vibrations in aero-engines as described above.

[0029] The present invention has the following beneficial effects:

[0030] 1. The present invention does not have high requirements for the number of engine data units, and is applicable to the development stage, new engine manufacturing and delivery stage, and maintenance delivery stage. It is also applicable to a small number of data units. By applying intelligent technology, an intelligent comparative analysis method suitable for excessive vibration is built, thereby establishing the influence relationship between excessive vibration parameters and assembly parameters, providing a reference for the replacement of parts in engines with excessive vibration. While saving manpower and resources in engine testing and manufacturing, it can accurately locate the root cause of vibration problems, shorten the product manufacturing cycle, and reduce the risk of delayed delivery.

[0031] 2. The present invention introduces environmental parameters, state parameters, control parameters, and vibration parameters, and constructs a vibration digital twin model based on the engine's physical structure and vibration physical mechanism. This makes the present invention more consistent with physical reality and avoids the underfitting problem caused by directly feeding assembly parameters into the neural network for training (due to insufficient assembly parameter data), thereby improving the model's accuracy and physical interpretability. Attached Figure Description

[0032] Figure 1 This is a flowchart of the method of the present invention;

[0033] Figure 2 This invention provides a vibration digital twin model.

[0034] Figure 3 This is another vibration digital twin model of the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Unless otherwise specified, the technical means used in the embodiments are conventional means well known to those skilled in the art.

[0036] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this invention, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.

[0037] like Figure 1-3 As shown, the digital twin method for troubleshooting vibration in aero-engines includes the following steps:

[0038] S1: Collect engine vibration test datasets. The vibration test dataset is a collection of test data from several engines. The data in the vibration test dataset is then filtered, analyzed, and processed. To achieve the purpose of test comparison and analysis, the vibration test dataset needs to include engine data with normal vibration as well as data with abnormal vibration to form the test dataset.

[0039] S2: Train the test data of each engine using the engine vibration digital twin model to obtain the vibration weight model of each engine; train the test data of each engine in step S1 using the engine vibration digital twin model.

[0040] S3: Calculate the accuracy of each vibration weight model; calculate the accuracy of each vibration weight model, where model accuracy refers to the residual value of the loss function or the prediction accuracy of the validation dataset.

[0041] S4: If the vibration weighting model accuracy is within the threshold range, proceed to the next step; if the vibration weighting model exceeds the threshold range, return to step S1 to check and analyze the data and retrain the model. If the accuracy is within the threshold range, the trained model is reliable; otherwise, if it is outside the threshold range, the reliability of the model is questionable, and the data needs to be checked, analyzed, and retrained. The threshold range is defined as the residual of the loss function being less than 0.001, or the prediction accuracy of the validation dataset being less than 1%.

[0042] S5: Substitute the ground test data of all engines into each vibration weight model in turn to conduct a vibration digitization test and obtain a vibration digitization test matrix; Substitute the data in step S1, specifically including the engine's environmental parameters, state parameters and control parameters, into different vibration weight models in turn to form a vibration digitization test matrix. The rows in this matrix represent the different vibration states corresponding to each engine, and the columns in this matrix represent different engines under the same vibration state.

[0043] S6: Analyze the vibration digitization test matrix in step S5 according to the vibration state of the engine, and identify the engines with excessive vibration.

[0044] S7: Compare the extreme values ​​of the assembly parameters and rotor balance parameters of the engine with excessive vibration to the corresponding extreme values ​​of the assembly parameters and rotor balance parameters of other engines to determine the location of the problematic part. Replace the problematic part and reassemble and test drive.

[0045] Example 1

[0046] This embodiment is based on the digital twin method for troubleshooting engine vibration described above, and will be explained in detail, assuming that the number of engines tested is n.

[0047] S1: Collect data on engines with abnormal vibration, and examine and analyze the data, removing abnormal data. To achieve the purpose and requirements of comparative analysis, data from engines with normal vibration also need to be added. These, together with the abnormal vibration data, form the test dataset. All data here needs to be filtered, analyzed, and processed. Specifically, the data filtering and processing methods include identifying abnormal data, including data with abnormal data types, null / infinite value data, etc., and removing or adding constant values ​​to these data.

[0048] Data analysis can be used to eliminate parameters that have low correlation with the target vibration value;

[0049] S2: The test data of each engine are trained using a digital twin model of engine vibration to obtain a vibration weight model for each engine. The final vibration weight models are M1,...,M n In this embodiment, a digital twin model of engine vibration is provided. The digital twin model of engine vibration includes a feature input layer, a coupling network layer, and a vibration mapping network layer. The coupling network layer extracts all parameters from the feature input layer, couples them, and then transmits them to the vibration mapping network layer. The feature input layer includes feature parameters input by various parts or components within the engine.

[0050] refer to Figure 2 The feature input layer of this model mainly includes the intake component network, fan component network, compressor component network, combustion chamber component network, turbine component network, and exhaust nozzle component network. The feature parameters of each component are mainly composed of environmental parameters, state parameters, and control parameters during the test run.

[0051] The feature parameters corresponding to the intake duct network are intake duct parameters, such as intake pressure and temperature; the feature parameters corresponding to the fan component network are fan parameters, such as low-pressure fan speed; the feature parameters corresponding to the compressor component network are compressor parameters, such as compressor guide angle and high-pressure speed; the feature parameters corresponding to the combustion chamber component network are combustion chamber parameters, such as fuel manifold pressure; the feature parameters corresponding to the low-pressure turbine component network are low-pressure turbine component parameters, such as low-pressure turbine speed; the feature parameters corresponding to the high-pressure turbine component network are low-pressure turbine component parameters, such as high-pressure turbine speed; and the feature parameters corresponding to the exhaust nozzle component network are exhaust nozzle component parameters, such as exhaust nozzle diameter or exhaust nozzle distance. The coupling network layer extracts the matching operational physical features of each component, while the mapping layer maps the component's operational physical features to the vibration values ​​of interest. The vibration values ​​during the test run mainly consist of the horizontal or vertical vibration values ​​of the intake casing, intermediate casing, and turbine rear casing.

[0052] S3: Calculate the accuracy of each vibration weight model; calculate the accuracy of each vibration weight model, where model accuracy refers to the residual value of the loss function or the prediction accuracy of the validation dataset.

[0053] S4: If the vibration weighting model accuracy is within the threshold range, proceed to the next step; if the vibration weighting model exceeds the threshold range, return to step S1 to check and analyze the data and retrain the model. If the accuracy is within the threshold range, the trained model is reliable; otherwise, if it is outside the threshold range, the reliability of the model is questionable, and the data needs to be checked, analyzed, and retrained. The threshold range is defined as the residual of the loss function being less than 0.001, or the prediction accuracy of the validation dataset being less than 1%.

[0054] S5: Substitute the ground test data of all engines into each vibration weighting model in turn to conduct a vibration digitization test and obtain the vibration digitization test matrix; specifically, substitute the environmental parameters, state parameters and control parameters of engine No. 1 into M1 to M... n All vibration models were obtained, resulting in the digital vibration test results of engines 1 to n under the test conditions of engine 1. Following the same approach, the environmental parameters, state parameters, and control parameters of engines 2 to n were sequentially substituted into M1 to M... n All vibration models are obtained, that is, the digital vibration test results of engines 1 to n under the test conditions of engines 2 to n are obtained.

[0055] Taking the digital vibration test of engines 1 to n under the test condition of engine 1 as an example, since the input data is consistent, namely the environmental parameters, state parameters and control parameters of engine 1, the test results are all obtained under the same level conditions. However, models M1 to M... n These represent the vibration levels of engines 1 to n, respectively. Therefore, the digital vibration test results characterize the vibration levels of engines 1 to n under the same environment, state, and control (here, the environment, state, and control of engine 1). In this case, statistical comparison of the results or visualization comparison of the results graphs has practical significance and value.

[0056] S6: Based on the n*n digital vibration test results in step S05, perform statistical comparative analysis or graphical visualization analysis to identify engines with excessive vibration.

[0057] S7: Calculate the extreme values ​​of all engine assembly parameters and rotor balance parameters, and the engine number corresponding to the extreme values. If the engine number is consistent with the engine number of the engine with excessive vibration, then mark the part corresponding to this parameter as the problem part.

[0058] Example 2

[0059] The difference from the above embodiments is that, in this embodiment, the engine vibration digital twin model described in step S2 is constructed in sequence into three vibration output sub-networks according to the physical operating mechanism or gas flow direction inside the engine. The vibration output sub-network includes a feature input layer, a coupling network layer, and a vibration mapping network layer; the feature input layer of the subsequent vibration output sub-network contains all the feature parameters in the feature input layer of the preceding vibration output sub-network.

[0060] refer to Figure 3 The network architecture strictly adheres to the physical operating mechanism of the engine, arranged according to its physical geometry. It follows the geometric structure or gas flow direction from the intake manifold—fan—compressor—combustion chamber—high-pressure turbine—low-pressure turbine—nozzle. Simultaneously, the vibration parameter-related network layers are consistent with the actual engine's measurement layout. For example, vibration sensors are placed on the intake casing, intermediate casing, and turbine rear casing to monitor the vibration of the engine's front, middle, and rear sections. Therefore, for... Figure 3 The vibration digital twin model is connected to the intake casing network, intermediate casing network, and turbine rear casing network at the outlet of the intake training network layer, the outlet of the combustion chamber network layer, and the outlet of the nozzle network layer, respectively.

[0061] The three vibration output sub-networks each include a feature input layer, a coupling network layer, and a vibration mapping network layer. For the feature input layer in the intermediate casing network, the input feature parameters include the parameters in the feature input layer of the intake casing network. For the feature input layer in the turbine rear casing network, the input feature parameters include the parameters in the feature input layers of the intake casing network and the intermediate casing network.

[0062] Example 3

[0063] This embodiment, based on either Embodiment 1 or Embodiment 2, provides a specific method for identifying engines with excessive vibration. Specifically, for step S6, the method is as follows: compare the vibration characteristic parameter values ​​obtained by each engine in the same vibration weighting model. If the vibration characteristic parameter value is greater than a set value, the engine is determined to be an engine with excessive vibration.

[0064] For the comparative evaluation of vibration, the main focus is on calculating and analyzing an n*n digitized vibration test matrix. The parameters used for analysis are primarily the vertical vibration of the intermediate casing and the vertical vibration behind the turbine, with the vertical vibration of the inlet casing serving as a secondary analysis variable. Assuming Z is a vibration characteristic parameter, it encompasses all vibration parameters from the inlet casing, intermediate casing, and turbine rear casing. ij This indicates that engine j is in vibration model M i The vibration quantity prediction results of the digital vibration test were obtained. Based on the aforementioned information, Z...1j This indicates the vibration level of engine No. 1 under test conditions from No. 1 to No. n. In other words, these vibration values ​​all represent the vibration level of engine No. 1, just under different test conditions.

[0065] To compare the vibration levels of engines 1 to n, the digital vibration results under the same conditions can be graphically visualized and compared, i.e., Z... 1j Z 2j ... Z n-1j Z nj By making comparisons and determining the vibration limits, we can identify the Z values ​​that exceed the limits. ij The corresponding engine number i is the engine with excessive vibration.

[0066] Example 4

[0067] The difference from Embodiment 3 is that this embodiment provides another method for identifying engines with excessive vibration, specifically including:

[0068] S6.1: Obtain several vibration characteristic parameter values ​​for each engine in the same vibration weighting model, and calculate the mean of each vibration parameter value; for example: calculate the mean vibration of each engine under the condition of engine number 1. These values ​​represent the vibration capabilities of engines 1-n, and are used as the vibration capability scores of engines 1-n under the test conditions of engine 1. The vibration capability scores of engines 1-n under the test conditions of engines 2-n are calculated in the same way.

[0069] S6.2: Summing the mean values ​​of vibration parameters obtained from various vibration weighting models for the same engine yields the engine's vibration score; the summation formula is:

[0070]

[0071] Similarly, the vibration scores S2, ..., S of engines 2-n are calculated. n S n The higher the score, the more severe the vibration and the easier it is to exceed the vibration standard. At the same time, S is a vector that contains the vibration scores of all vibration parameters, such as the vertical vibration of the intake casing, the vertical vibration of the intermediate casing, and the vertical vibration of the turbine rear casing.

[0072] S6.3: Determine whether the vibration score of each engine reaches or exceeds the set value. If so, it is considered an engine with excessive vibration. Set a score threshold. For example, if the limit value of vibration parameter Z is L, then n*L is set as the score threshold. Engines exceeding this value are defined as engines with excessive vibration.

[0073] The above uses a limit value approach, but a standard engine approach can also be used. For example, if engine No. 1 is designated as a standard engine with normal vibration, then any engine that exceeds the vibration value of engine No. 1 under the same conditions can be defined as an engine with excessive vibration.

[0074] Example 5

[0075] This embodiment, based on any combination of Embodiment 1, Embodiment 2, and Embodiments 3 and 4, provides a method for determining the location of a problematic part. Specifically, the method involves comparing the assembly parameters and rotor balance parameters of the engine with excessive vibration with the assembly parameters and rotor balance parameters of all other engines. If the assembly parameters and rotor balance parameters of the engine with excessive vibration are the maximum, then the part corresponding to the parameters is considered to be the problematic part in the engine with excessive vibration.

[0076] The vibration scores, assembly parameters, and rotor balance parameters of engines 1-n are displayed as curves on the same graph. By comparison, abnormal assembly parameters and rotor balance parameters in the engines that exceed the standard are identified. The parts at the location of these parameters are the problematic parts. If there are assembly parameters and rotor balance parameters A, B, C, and D, and parameter A of the engine that exceeds the standard has the largest value among all engines, then the part corresponding to parameter A is the problematic part of the engine that exceeds the standard.

[0077] Assembly parameters refer to the bearing clamping nut torque, bearing inner and outer ring fit clearance, coaxiality, end face runout value, cylindrical surface runout value, etc. (each may have more than one parameter, and may include parameters for multiple positions). Rotor balance parameters mainly refer to the balance parameters of rotor components, which mainly refer to fans, compressors, low-pressure turbines, and high-pressure turbines.

[0078] Example 6

[0079] The difference from Embodiment 5 above lies in that this embodiment provides a different method for identifying problematic parts. Specifically, the method involves using the Pearson correlation coefficient to calculate the correlation between vibration parameters, assembly parameters, and rotor balance parameters. The assembly parameters and rotor balance parameters that result in high vibration scores are identified, and the parts corresponding to these parameters are the problematic parts. The calculation formula is as follows:

[0080]

[0081] In the above formula, the numerator is the covariance of the two parameters X and Y. The denominators are the standard deviations of the two parameters X and Y, respectively, and the result is the correlation coefficient between the two parameters X and Y.

[0082] Based on the above formula, the Pearson correlation coefficient between vibration parameters, assembly parameters, and rotor balance parameters can be calculated. The closer the absolute value of the Pearson correlation coefficient is to 1, the higher the correlation; the closer it is to 0, the almost no correlation. This method can screen out assembly parameters and rotor balance parameters that are highly correlated with vibration scores. However, it is still impossible to determine whether these parameters are the main cause of high vibration scores. Therefore, interpolation can be used to calculate the linear numerical relationship between these parameters and vibration scores, and the assembly parameters and rotor balance parameters that are proportional to vibration scores can be obtained through the relationship formula. Here, the vibration parameter used for correlation analysis is the vibration parameter. By using the above method, the assembly parameters and rotor balance parameters that cause high vibration scores can be found. The parts corresponding to these parameters are the problematic parts, which can be replaced, reassembled, and retested.

[0083] The present invention also provides an electronic device, comprising:

[0084] One or more processors;

[0085] Storage device for storing one or more programs;

[0086] When the one or more programs are executed by the one or more processors, the one or more processors implement the intelligent analysis method for vibration faults of aero-engines as described above.

[0087] The present invention also provides a computer-readable medium storing a computer program that, when executed by a processor, implements the intelligent analysis method for vibration faults of aero-engines as described above.

[0088] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Any modifications, alterations, substitutions, or variations made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention shall fall within the protection scope defined by the claims of the present invention.

Claims

1. A digital twin method for aeroengine vibration troubleshooting, characterized in that, The method comprises the following steps: S1: collecting a vibration test data set of the engine, the vibration test data set being a collection of test data of a plurality of engines, and screening, analyzing and processing the data in the vibration test data set; S2: training the test data of each engine using an engine vibration digital twin model to obtain a vibration weight model of each engine; S3: calculating the accuracy of each vibration weight model; S4: if the vibration weight model accuracy is within the threshold range, proceed to the next step; if the vibration weight model accuracy exceeds the threshold range, return to step S1 to check, analyze and process the data and retrain the model; S5: sequentially input the ground test data of all engines into each vibration weight model to perform vibration digital test and obtain a vibration digital test matrix; S6: analyzing the vibration digital test matrix in step S5 according to the vibration state of the engine to find out the engine with excessive vibration; S7: comparing the extreme values of the assembly parameters and rotor balance parameters of the engine with excessive vibration with the extreme values of the assembly parameters and rotor balance parameters of other engines to determine the location of the problem part.

2. The digital twin method for aero-engine vibration troubleshooting according to claim 1, characterized in that: The engine vibration digital twin model in step S2 comprises a feature input layer, a coupling network layer and a vibration mapping network layer; the coupling network layer extracts all parameters of the feature input layer for coupling and then transmits them to the vibration mapping network layer; the feature input layer comprises characteristic parameters input by each part or component in the engine.

3. The digital twin method for aero-engine vibration troubleshooting according to claim 1, characterized in that: The engine vibration digital twin model in step S2 sequentially forms three vibration output sub-networks according to the physical operation mechanism or gas flow direction of the engine, wherein each vibration output sub-network comprises a feature input layer, a coupling network layer and a vibration mapping network layer; the feature input layer of the rear vibration output sub-network comprises all characteristic parameters in the feature input layer of the front vibration output sub-network.

4. The digital twin method for aero-engine vibration troubleshooting according to any one of claims 1-3, characterized in that: The specific method for finding out the engine with excessive vibration in step S6 is to compare the values of a plurality of vibration characteristic parameters obtained for each engine in the same vibration weight model, and if the value of a vibration characteristic parameter is greater than a set value, the engine is determined to be an engine with excessive vibration.

5. The digital twin method for aero-engine vibration troubleshooting according to any one of claims 1-3, characterized in that: The specific method for finding out the engine with excessive vibration in step S6 is: S6.1: obtaining a plurality of vibration characteristic parameter values for each engine in the same vibration weight model, and calculating the mean value of each vibration parameter value; S6.2: summing the mean values of the vibration parameter values obtained for the same engine in each vibration weight model to obtain the vibration score of the engine; S6.3: determining whether the vibration score of each engine reaches or exceeds a set value, and if so, considering the engine to be an engine with excessive vibration.

6. The digital twin method for aero-engine vibration troubleshooting according to claim 5, characterized in that: In step S7, the specific method for determining the location of the problem part is to compare the assembly parameters and rotor balance parameters of the engine with excessive vibration with the assembly parameters and rotor balance parameters of all other engines; if the assembly parameters and rotor balance parameters of the engine with excessive vibration are the maximum, the part corresponding to the parameters is considered to be the problem part in the engine with excessive vibration.

7. The digital twin method for aero-engine vibration troubleshooting according to claim 5, characterized in that: In step S7, the specific method for determining the location of the problem part is to calculate the correlation degree between the vibration parameters and the assembly parameters and the rotor balance parameters by using the Pearson correlation coefficient, find the assembly parameters and the rotor balance parameters that cause high vibration scores, and the parts corresponding to the parameters are the problem parts.

8. The digital twin method for aero-engine vibration troubleshooting according to claim 6 or 7, characterized in that: The assembly parameters include bearing compression nut torque, bearing inner and outer ring fit clearance, coaxiality, end face runout value, and cylindrical surface runout value; and the rotor balance parameters include balance parameters of the fan, the compressor, the low-pressure turbine, and the high-pressure turbine.

9. An electronic device, comprising: Comprising: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the digital twin method for aero-engine vibration troubleshooting according to any one of claims 1-7.

10. A computer readable medium, said readable medium storing a computer program, characterized in that: The computer program is executed by the processor to implement the digital twin method for aero-engine vibration troubleshooting according to any one of claims 1-7.

Images