Method, device and equipment for predicting service life of high-pressure turbine blade in aero-engine

Abstract

The invention relates to the technical field of risk detection, and discloses a method, device and equipment for predicting the service life of a high-pressure turbine blade in an aero-engine, and the method comprises the steps: carrying out the statistics of multi-dimensional factor parameters related to the aero-engine, and decomposing the multi-dimensional factor parameters into a plurality of feature parameter groups; acquiring a data sample set of the feature parameter group, and selecting a parameter model adaptive to the feature parameter group; training and verifying the parameter models by utilizing the data sample set, and adding the parameter models of which verification results meet preset conditions into a model cache pool; screening a representative model from the model cache pool, and performing multi-dimensional prediction on actual data of the target engine by using the representative model; a risk prediction level for the target engine is generated. According to the technical scheme, the risk level of the engine can be accurately predicted, and then the service life of the high-pressure turbine blade is obtained.

Description

Technical Field

[0001] The present application relates to the field of risk detection technology, and in particular to a method, device and equipment for predicting the life of high-pressure turbine blades in an aircraft engine. Background Art

[0002] Civil aviation engines are composed of a fan, a low-pressure compressor, a high-pressure compressor, a combustion chamber, a high-pressure turbine, and a low-pressure turbine. After the gas is mixed with the fuel in the combustion chamber and expanded and burned, it first passes through the high-pressure turbine (HPTB), which drives the high-pressure compressor in front, and then passes through the low-pressure turbine, which drives the low-pressure compressor and the fan. It can be said that the HPTB is a more critical and core power source for the engine. Once it is damaged, the high-pressure compressor cannot work, the engine will lose all power, and it will cause an air shutdown.

[0003] Conventional civil aviation engines are inspected using a borescope, which uses an industrial endoscope to reach into the engine and display the image through a camera for inspection by professional engineers. However, the key part of the HPTB component is the bottom connection tenon, which cannot be directly seen and cannot be inspected using a borescope.

[0004] Therefore, with current technology, it is impossible to detect the root status of the high-pressure turbine blades when the engine is on-wing, and the life of the components can only be controlled by relying on the service life (number of flight cycles) provided by the component manufacturer. Obviously, such a control method cannot accurately know the working status of the high-pressure turbine blades, and there may be a risk of downtime in the air. Summary of the invention

[0005] The present application provides a life prediction method, device and equipment for high-pressure turbine blades in an aircraft engine, which can accurately predict the risk level of the engine without disassembling the engine, and thus know the service life of the high-pressure turbine blades.

[0006] In a first aspect, an embodiment of the present application provides a method for predicting the life of a high-pressure turbine blade in an aircraft engine, the method comprising: statistically analyzing multidimensional factor parameters related to the aircraft engine, and decomposing the multidimensional factor parameters into multiple characteristic parameter groups, the multidimensional factor parameters are used to characterize engine performance parameters and / or flight status parameters, the characteristic parameter group comprising one or more factor parameters; for any characteristic parameter group, obtaining a data sample set of the characteristic parameter group, and selecting a parameter model adapted to the characteristic parameter group according to the number of factor parameters included in the characteristic parameter group; wherein, if the number of factor parameters included in the characteristic parameter group is more than one, the selected parameter model The method comprises at least a parallel network branch for performing weight splitting; using the data sample set to train and verify the parameter model, and adding the parameter model whose verification result meets the preset conditions to the model cache pool; selecting respective representative models for each characteristic parameter group from the model cache pool, and using each selected representative model to perform multi-dimensional prediction on the actual data of the target engine, respectively, to obtain multiple single-dimensional prediction results; based on the weight coefficient assigned to each of the characteristic parameter groups, weighted summing the multiple single-dimensional prediction results is performed to generate a risk prediction level for the target engine, wherein the risk prediction level is used to characterize the service life of the high-pressure turbine blades in the target engine.

[0007] The technical solution provided in this embodiment can decompose the multidimensional factor parameters related to the aircraft engine into multiple characteristic parameter groups. By training the model of each characteristic parameter group, the parameter model of each characteristic parameter group can be obtained. During the training process, according to the number of factor parameters contained in the characteristic parameter group, parameter models with different structures can be selected to make the feature processing process more accurate. Subsequently, by screening out the representative models of each characteristic parameter group, different single-dimensional prediction results can be predicted using the representative models. By weighted summing these single-dimensional prediction results, an accurate risk prediction level can be obtained. The risk prediction level can be used to characterize the service life of high-pressure turbine blades. Through the technical solution provided in this embodiment, the risk level of the engine can be accurately predicted without disassembling the engine, and the service life of the high-pressure turbine blades can be known, effectively avoiding the situation of air shutdown.

[0008] In one embodiment, the parameter model is trained and verified using the data sample set, and the parameter model whose verification result meets the preset conditions is added to the model cache pool, including: extracting a first proportion of data from the data sample set as a global verification set, wherein the global verification set includes data samples representing a healthy state and data samples representing an unhealthy state; performing specified batches of training on the parameter model, wherein each batch of training process includes a specified number of groups of iterative training, and each group of iterative training includes a specified number of rounds of training; wherein, for any batch of training process, a second proportion of data is selected from a data sample set other than the global verification set as a local verification data set, and in any group of iterative training, a third proportion of data is selected as a local verification data set. as a test data set, and selecting a fourth proportion of data as a training data set; in each round of training, using the training data set to train the parameter model, and after completing the training, using the test data set to test the trained parameter model to determine the direction of gradient descent; after completing all rounds of training and testing in a set of iterative training, reselecting the test data set and the training data set, and entering the next set of iterative training processes based on the reselected test data set and training data set; after completing all groups of iterative training in a batch, using the local verification data set to verify the trained parameter model, and adding the parameter model with a verification accuracy greater than or equal to the specified accuracy threshold to the model cache pool, and entering the next batch of training processes.

[0009] The technical solution provided in this embodiment can realize a relatively complete model training process by setting different batches, numbers of groups and numbers of rounds. In the training process, the efficiency and accuracy of the training process can be guaranteed by setting a global verification set, a local verification data set, a test data set and a training data set.

[0010] In one embodiment, the method further includes: if the verification accuracy is less than a specified accuracy threshold, the parameter model obtained by the current batch training is discarded, and the training process of the next batch is entered.

[0011] In one embodiment, the first ratio is 5%, the second ratio is 5%, the third ratio is 10%, the fourth ratio is 80%, the designated batch is 100 times, the designated number of groups is 15 groups, and the designated number of rounds is 30 rounds.

[0012] In one embodiment, the weight coefficient of each of the feature parameter groups is determined in the following manner: for any feature parameter group, the number of parameter models possessed by the feature parameter group in the model cache pool is identified, and the training batches of the parameter models of the feature parameter group during the training process are determined; the ratio between the number of the parameter models and the training batches is calculated, and the ratio is used as the model fitting parameter of the feature parameter group; the proportion of the model fitting parameters of the feature parameter group in the total model fitting parameters of each feature parameter group is calculated, and the proportion is used as the weight coefficient of the feature parameter group.

[0013] The technical solution provided in this embodiment utilizes the number of parameter models and training batches in the model cache pool to obtain relatively accurate model fitting parameters, and then obtain relatively accurate weight coefficients, providing an accurate data basis for the subsequent generation of the final risk prediction level.

[0014] In one embodiment, in the process of training and verifying the parameter model using the data sample set, the parameter model processes the data samples in the data sample set in the following manner: if the number of factor parameters contained in the feature parameter group is 1, the data sample is normalized, and a three-layer fully connected network is used to perform feature classification on the normalized features, wherein an L2 regularization mechanism is added to each layer of the fully connected network for correction, and an early stopping mechanism is added to prevent overfitting. In terms of the loss function, a customized loss function is defined based on the ReLU function, and the loss function adds a penalty for the situation where the actual sample is in an unhealthy state but the predicted result is in a healthy state; if the number of factor parameters contained in the feature parameter group is more than 1 The weights of the parameter model are split, wherein the first part of the weights after the split enters the random forgetting branch, and the second part of the weights enters the feature processing branch, wherein the feature processing branch includes a convolutional layer, a batch normalization layer, and a pooling layer; after the weights output by the random forgetting branch and the feature processing branch are integrated, they enter the attention layer for dynamic weight allocation; a three-layer fully connected network is used to classify the features after the dynamic weight allocation, wherein an L2 regularization mechanism is added to each layer of the fully connected network for correction, and an early stopping mechanism is added to prevent overfitting. In terms of the loss function, a customized loss function is defined based on the ReLU function, and the loss function adds a penalty for the situation where the actual sample is in an unhealthy state but the predicted result is in a healthy state.

[0015] The technical solution provided in this embodiment can adopt parameter models of different structures for different numbers of factor parameters. When the feature parameter group contains only a single factor parameter, a three-layer fully connected network can be used to process the input sample and obtain the corresponding prediction result. When the feature parameter group contains multiple factor parameters, since it involves the fusion of factor parameters, the perceptual field of the model can be gradually increased by first splitting the weights and then integrating the weights, so that more attention can be paid to the overall situation of the sample, so that the model can better predict risks based on multiple factor parameters. Subsequently, by dynamically allocating weights through the attention layer, more accurate fusion features can be obtained, and based on the accurate fusion features, accurate risk prediction levels can be obtained.

[0016] In one embodiment, selecting respective representative models for respective feature parameter groups from the model cache pool includes: for any feature parameter group, verifying the parameter model corresponding to the feature parameter group in the model cache pool using a preset global verification set, and determining the parameter model as a candidate parameter model if the parameter model correctly predicts the verification data characterizing the idle stop state in the global verification set; obtaining the verification result of the candidate parameter model for the global verification set, and generating the cumulative residual of the candidate parameter model according to the verification result when the verification result meets a preset verification condition, and determining the candidate parameter model with the smallest cumulative residual as the representative model of the feature parameter group.

[0017] In one embodiment, the verification result satisfies the preset verification conditions, including: for each verification sample in the global verification set, the risk type represented by the verification result is consistent with the pre-marked risk type, and the risk type includes risk and no risk; and the error between the risk prediction score represented by the verification result and the pre-marked risk standard score is within a specified range.

[0018] The technical solution provided in this embodiment uses a global validation set to verify the parameter model in the model cache pool, and can screen out a representative model that meets the requirements. The representative model can accurately predict the risk type of the validation sample. At the same time, for more detailed risk prediction scores, it can also meet the error requirements, thereby achieving a higher prediction accuracy.

[0019] In one embodiment, the method further includes: if the score representing the risk prediction level of the target engine exceeds a preset score threshold within a specified time period, disassembling the high-pressure turbine blades in the target engine and obtaining parameter data of the disassembled high-pressure turbine blades; and adding the obtained parameter data to the global verification set and the training data set respectively.

[0020] The technical solution provided in this embodiment is that if the risk prediction score of a target engine is high, there will be a risk of in-flight shutdown. At this time, the target engine can be disassembled and the parameter data of the high-pressure turbine blades obtained from the disassembly can be fed back to the HPTB's full life cycle database, thereby providing data support for the training data set and the global validation set, which is conducive to the subsequent training of a more accurate parameter model.

[0021] In a second aspect, the present application further provides a life prediction device for a high-pressure turbine blade in an aircraft engine, the device comprising:

[0022] A feature decomposition unit, used for counting multidimensional factor parameters related to the aircraft engine, and decomposing the multidimensional factor parameters into a plurality of feature parameter groups, wherein the multidimensional factor parameters are used to characterize engine performance parameters and / or flight status parameters, and the feature parameter group includes one or more factor parameters;

[0023] A model adaptation unit, for obtaining a data sample set of any characteristic parameter group, and selecting a parameter model adapted to the characteristic parameter group according to the number of factor parameters included in the characteristic parameter group; wherein, if the number of factor parameters included in the characteristic parameter group is more than one, the selected parameter model at least includes a parallel network branch for performing weight splitting;

[0024] A model training unit, used to train and verify the parameter model using the data sample set, and add the parameter model whose verification result meets the preset conditions to the model cache pool;

[0025] A prediction unit, used to select respective representative models for respective characteristic parameter groups from the model cache pool, and use the selected representative models to perform multi-dimensional prediction on the actual data of the target engine to obtain multiple single-dimensional prediction results;

[0026] A risk assessment unit is used to perform weighted summation on the multiple single-dimensional prediction results based on the weight coefficients assigned to each of the characteristic parameter groups to generate a risk prediction level for the target engine, wherein the risk prediction level is used to characterize the service life of the high-pressure turbine blades in the target engine.

[0027] In a third aspect, the present application also provides a computer device, comprising a memory and a processor, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the above-mentioned method for predicting the life of high-pressure turbine blades in an aircraft engine is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the specific implementation methods of the present application or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0029] Figure 1 A step diagram of a life prediction method for a high-pressure turbine blade in an aircraft engine provided in an embodiment of the present application;

[0030] Figure 2 A schematic diagram of a process for predicting the life of a high-pressure turbine blade in an aircraft engine in one embodiment of the present application;

[0031] Figure 3 A specific processing flow chart of the parameter model in an application scenario for this application;

[0032] Figure 4 A schematic diagram of functional modules of a life prediction device for a high-pressure turbine blade in an aircraft engine provided by one embodiment of the present application;

[0033] Figure 5 A schematic diagram of the structure of a computer device provided for one embodiment of the present application. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0035] In addition, the descriptions involving "first", "second", etc. in this application are for descriptive purposes only and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Thus, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the embodiments of the present application, unless otherwise specified, "multiple" means two or more. In addition, the use of "based on" or "according to" means openness and inclusiveness, because the process, steps, calculations or other actions "based on" or "according to" one or more of the conditions or values ​​can be based on additional conditions or beyond the described values ​​in practice.

[0036] During the operation of the aircraft, all data generated by sensors will be recorded in a certain format, called QAR (Quick access recorder) data. This data will be automatically uploaded to the cloud after the aircraft lands. Engineers can use QAR data to restore a series of information such as flight status, engine operating conditions, status, etc. This data can be used as a basis for fault diagnosis. QAR data has different recording frequencies, which will increase with the importance of the sensor whose parameters are recorded, and is usually recorded once every 1 second.

[0037] Currently, airlines control flight data based on the tail number of the aircraft. In actual work, parts (including engines) are sometimes replaced or installed on other aircraft. When it is necessary to track and control parts, the data is confusing and difficult to trace completely. This is obviously not ideal for parts control.

[0038] In view of this situation, this application proposes a component-based full life cycle database. This management method is component-centric and establishes a cloud-based electronic database. Starting from the first installation of the component, it records: installation time, installation machine tail number, installation location and other information. Each time the component is subjected to relevant work, additional records are made: removal time, removal reason, work performed and other information, so as to facilitate data management and traceability of the entire life cycle of the engine HPTB model.

[0039] In practical applications, this application provides a feature extraction method for aircraft engine QAR data, which can quickly extract the accumulated operating conditions of the engine from massive QAR historical data. Specifically, different Labels can be used to correspond to different risk levels, and each risk level can correspond to a different score interval. The correspondence between Label and score interval can be used for subsequent model training. In a specific application example, the score intervals corresponding to different Labels can be shown in Table 1.

[0040] Table 1 Score intervals corresponding to different Labels

[0041]

[0042] In this embodiment, the risk level of the Label may correspond to the severity of the HPTB crack failure. Specifically, the risk level may be as follows:

[0043] Level 7: The HPTB breaks into at least two parts along the small neck, and the turbine disk cannot prevent the part above the small neck from detaching from the turbine disk. The broken part can completely detach from the engine (fly out).

[0044] Level 6: The HPTB breaks into two parts along the small neck, but the remaining material can still be restrained by the turbine disk and will not fall out;

[0045] Level 5: There are cracks along the small neck of the HPTB that extend to the end surface, and the cracks extend to the third surface (for example: the initiation surface, the end surface, the bottom surface or another side surface);

[0046] Level 4: There are cracks along the small neck of the HPTB that extend to the end surface and extend to more than 1mm inside the end surface, but the bottom does not extend to the third surface (for example: the initiation surface, the end surface, the bottom surface or another side surface);

[0047] Level 3: There are cracks extending from the small neck of HPTB to the end face, but the extension length in the end face is less than 1mm;

[0048] Level 2: HPTB has cracks in the pressure surface area, and the FPI indicates that the crack length is greater than 2 mm, but does not extend to the end face;

[0049] Level 1: HPTB has tiny cracks in the pressure surface area, and the FPI indicates that the crack length is less than 2 mm, which is defined as the crack initiation stage;

[0050] Level 0: HPTB does not have any cracks in the pressure surface area.

[0051] Among them, level 1 is the crack initiation period, levels 1-3 are the crack growth period, levels 3-4 are the rapid development period, levels 4-6 are the dangerous outbreak period, and level 7 is the failure period.

[0052] According to the state of the HPTB when the engine is removed, the HPTB below level 2 (including level 2) can be defined as a healthy state, and the HPTB above level 3 (including level 3) can be defined as an unhealthy state.

[0053] See also Figure 1 and Figure 2 An embodiment of the present application provides a life prediction method for a high-pressure turbine blade in an aircraft engine, which may include the following steps.

[0054] S1: Counting multidimensional factor parameters related to aircraft engines, and decomposing the multidimensional factor parameters into multiple characteristic parameter groups, wherein the multidimensional factor parameters are used to characterize engine performance parameters and / or flight status parameters, and the characteristic parameter group includes one or more factor parameters.

[0055] In this embodiment, the multidimensional factor parameters related to the aircraft engine can be engine performance parameters or flight state parameters. For example, in practical applications, the factor parameters can be engine performance parameters such as N2 speed and EGT (Exhaust Gas Temperature), and can also be flight state parameters such as climb rate and climb time. These factor parameters can respectively characterize the different states experienced by the aircraft engine. In this application, the multidimensional factor parameters can be decomposed into multiple characteristic parameter groups, and then a respective parameter model can be trained for each characteristic parameter group.

[0056] Specifically, the decomposed characteristic parameter group may include one or more factor parameters. For example, the N2 speed can be a characteristic parameter group alone, and the N2 parameter and the engine vibration parameter can also be a characteristic parameter group together. In other words, different characteristic parameter groups may contain different numbers of factor parameters. By independently training the model for each characteristic parameter group, the failure mechanism can be made explicit to a certain extent, laying the foundation for subsequent life prediction.

[0057] S2: For any feature parameter group, obtain a data sample set of the feature parameter group, and select a parameter model adapted to the feature parameter group according to the number of factor parameters included in the feature parameter group; wherein, if the number of factor parameters included in the feature parameter group is more than one, the selected parameter model includes at least a parallel network branch for performing weight splitting.

[0058] In this embodiment, an independent parameter model can be trained for each characteristic parameter group. Before training, a data sample set of the characteristic parameter group can be obtained. Specifically, each data sample in the data sample set is consistent with the type of the characteristic parameter group. For example, if the characteristic parameter group only contains a single climbing rate, then the data sample should also be the climbing rate. For another example, if the characteristic parameter group contains two factor parameters, the climbing rate and the climbing time, then the data sample should also contain the two factor parameters, the climbing rate and the climbing time. For each data sample, it can have its own label, which can be used to characterize the true risk score.

[0059] In this embodiment, according to the number of factor parameters included in the characteristic parameter group, parameter models with different structures can be selected for the characteristic parameter group. Figure 3 , if the feature parameter group contains only a single factor parameter, then the selected parameter model can be executed Figure 3The data processing flow on the left side of . Specifically, if the number of factor parameters contained in the feature parameter group is 1, the data sample can be normalized first, and a three-layer fully connected network is used to perform feature classification on the normalized features, wherein an L2 regularization mechanism is added to each layer of the fully connected network for correction, and an early stopping mechanism is added to prevent overfitting. In terms of the loss function, a custom loss function is defined based on the ReLU function, and the loss function adds a penalty for the case where the actual sample is in an unhealthy state but the predicted result is in a healthy state, so as to ensure that the model can better capture faults. Specifically, when calculating the loss residual, when the actual sample is in an unhealthy state but the interval where the predicted score is located is lower than the actual sample by 2 or more intervals, a correction factor α (α>1) is introduced. By multiplying the loss residual by α, the result focuses more on how to correctly predict unhealthy actual samples. Among them, α is a variable, which increases as the loss residual increases.

[0060] If the number of factor parameters contained in the feature parameter group is more than 1, the selected parameter model can be executed Figure 3 The processing process shown on the right side of . Specifically, if the number of factor parameters contained in the feature parameter group is more than one, after normalization, the weights of the parameter model can be split, wherein the first part of the weights after the split enters the random forgetting branch, and the second part of the weights enters the feature processing branch, wherein the feature processing branch includes a convolution layer, a batch normalization layer, and a pooling layer; after the weights output by the random forgetting branch and the feature processing branch are integrated, they enter the attention layer for dynamic weight allocation. In practical applications, the process of weight splitting and weight integration can be repeated three times. After entering the attention layer for dynamic weight allocation, similar to the processing process of a single factor parameter, a three-layer fully connected network can be used to classify the features after the dynamic weight allocation, wherein an L2 regularization mechanism is added to each layer of the fully connected network for correction, and an early stopping mechanism is added to prevent overfitting. In terms of loss function, a custom loss function is defined based on the ReLU function, and the loss function adds a penalty for the case where the actual sample is in an unhealthy state but the predicted result is in a healthy state. The weight splitting process can be regarded as dividing the input data into two parts: x1 and x2. One part passes through the convolution layer, batch normalization layer and pooling layer to obtain F(x2), and the other part passes through the random forgetting branch to obtain x1'. The final output is y=x1'+F(x2), so as to ensure that the gradient will not disappear or explode.

[0061] The technical solution provided in this embodiment can adopt parameter models of different structures for different numbers of factor parameters. When the feature parameter group contains only a single factor parameter, a three-layer fully connected network can be used to process the input sample and obtain the corresponding prediction result. When the feature parameter group contains multiple factor parameters, since it involves the fusion of factor parameters, the weight splitting can be performed first, and then the weight integration can be performed. This process can be carried out three times, the convolution kernel gradually increases, and the receptive field will gradually increase, so that the model can pay more attention to the overall situation of the sample, and can better predict risks based on multiple factor parameters. Subsequently, by dynamically allocating weights through the attention layer, a more accurate fusion feature can be obtained, and based on the accurate fusion feature, an accurate risk prediction level can be obtained.

[0062] S3: Using the data sample set to train and verify the parameter model, and adding the parameter model whose verification results meet the preset conditions to the model cache pool.

[0063] In this embodiment, after the data sample set is obtained and the corresponding parameter model is selected for the characteristic parameter group, the parameter model can be trained and verified using the data sample set.

[0064] In practical applications, different batches, different numbers of groups, and different rounds of training processes can be set. Specifically, a first proportion of data is extracted from the data sample set as a global validation set, and the global validation set includes data samples representing a healthy state and data samples representing an unhealthy state.

[0065] The parameter model is trained in designated batches, wherein the training process of each batch includes a designated number of iterative trainings, and each group of iterative training includes a designated number of rounds of training; wherein, for any batch of training, a second proportion of data is selected from a data sample set other than the global validation set as a local validation data set, and in any group of iterative training, a third proportion of data is selected as a test data set, and a fourth proportion of data is selected as a training data set.

[0066] In each round of training, the parameter model is trained using the training data set, and after the training is completed, the trained parameter model is tested using the test data set to determine the direction of gradient descent.

[0067] After completing all rounds of training and testing in a set of iterative training, reselect the test data set and the training data set, and enter the next set of iterative training process based on the reselected test data set and the training data set.

[0068] After completing the iterative training of all groups in a batch, the trained parameter model is verified using the local verification data set, and the parameter model with a verification accuracy greater than or equal to the specified accuracy threshold is added to the model cache pool, and the training process of the next batch is entered. If the verification accuracy is less than the specified accuracy threshold, the parameter model obtained by the current batch training is discarded, and the training process of the next batch is entered.

[0069] In a specific application example, the first ratio is 5%, the second ratio is 5%, the third ratio is 10%, the fourth ratio is 80%, the specified batch is 100 times, the specified number of groups is 15 groups, and the specified number of rounds is 30 rounds. The above training process can be described as follows:

[0070] Initial work: From the engines covered by the data sample set, pick out 5% of them and do not participate in any training. Only do global accuracy verification as the above-mentioned global verification set. In the global verification set, the number of healthy and unhealthy data samples is equal.

[0071] When executing the first batch of training, 5% of the engines covered by the data sample set (not repeated with the global validation set) are selected as the local validation data set for the first batch, and the local validation data set does not participate in the training.

[0072] In the first batch of training, data from 10% of the engines covered by the data sample set are selected as the test data set, and the data from the remaining 80% of the engines are used as the training data set. The training data set is put into the parameter model for training. After each round, the parameter model is tested with the test data set to determine the direction of gradient descent, and a total of 30 rounds of training are performed. After 30 rounds, the test data set and training data set are selected again (the proportion is the same as before, only the data screening is performed again), and the next group of training is performed again, for a total of 15 groups of training.

[0073] After completing the above process, that is, completing 15*30 rounds of training, verification work is performed. The parameter model that has completed training is verified using the selected local verification data set. If the verification accuracy is greater than or equal to 0.8 (the verification result meets the preset conditions), the parameter model that has completed training is saved to the model cache pool, otherwise the parameter model of this training is not saved. After a batch of training is completed, regardless of whether the parameter model is saved to the model cache pool, the next batch of training is performed.

[0074] Specifically, the verification accuracy can be calculated using the following formula:

[0075] TPVariance=(label i -predict i)2

[0076]

[0077] Among them, TPVariance represents the error variance, label i represents the true risk score of the ith validation sample, predict i represents the predicted risk score of the i-th validation sample, acc represents the validation accuracy, and n represents the total number of validation samples in the global validation set.

[0078] The technical solution provided in this embodiment can realize a relatively complete model training process by setting different batches, numbers of groups and numbers of rounds. In the training process, the efficiency and accuracy of the training process can be guaranteed by setting a global verification set, a local verification data set, a test data set and a training data set.

[0079] In one embodiment, after completing the model training for each feature parameter group, a weight coefficient can be set for each feature parameter group based on the number of parameter models retained in the model cache pool. Specifically, for any feature parameter group, the number of parameter models possessed by the feature parameter group in the model cache pool is identified, and the training batch of the parameter models of the feature parameter group during the training process is determined. The ratio between the number of parameter models and the training batch is calculated, and the ratio is used as the model fitting parameter of the feature parameter group. The proportion of the model fitting parameters of the feature parameter group in the total model fitting parameters of each feature parameter group is calculated, and the proportion is used as the weight coefficient of the feature parameter group.

[0080] For example, after training for one hundred batches of the N2 speed dimension:

[0081] The model cache pool stores k models of the N2 speed dimension, and trains a total of m batches (here 100). Then the model fitting parameter R of the N2 speed dimension is i It can be expressed as:

[0082]

[0083] After all feature parameter groups are trained, the total model fitting parameter R can be expressed as:

[0084] R=∑R

[0085] Then the weight coefficient w corresponding to the N2 speed dimension i It is expressed as:

[0086]

[0087] Finally, any w iShould be a number between 0 and 1, w i The larger it is, the easier it is for the model to converge, and the larger the corresponding weight is; i The smaller it is, the more difficult it is to correctly classify the results based on the current feature parameter group alone. Therefore, the models with verification accuracy greater than 0.8 are accidental phenomena caused by overfitting, and the corresponding weights are smaller.

[0088] The technical solution provided in this embodiment utilizes the number of parameter models and training batches in the model cache pool to obtain relatively accurate model fitting parameters, and then obtain relatively accurate weight coefficients, providing an accurate data basis for the subsequent generation of the final risk prediction level.

[0089] S4: Selecting respective representative models for respective characteristic parameter groups from the model cache pool, and using the selected representative models to perform multi-dimensional prediction on the actual data of the target engine, respectively, to obtain multiple single-dimensional prediction results.

[0090] In this embodiment, after completing the training of all batches, the parameter models corresponding to each feature parameter group can be saved in the model cache pool, and generally speaking, since each feature parameter group has been trained in a specified batch, the number of parameter models corresponding to the feature parameter group in the model cache pool is also relatively large. In practical applications, when using parameter models for risk prediction, the representative model of each feature parameter group can be screened out, and the representative model can be used to perform risk prediction of each dimension.

[0091] Specifically, for any feature parameter group, a pre-set global verification set can be used to verify the parameter model corresponding to the feature parameter group in the model cache pool. If the parameter model correctly predicts the verification data representing the idle stop state in the global verification set, the parameter model is determined as a candidate parameter model.

[0092] Then, the verification result of the candidate parameter model for the global verification set can be obtained. When the verification result meets the preset verification conditions, the cumulative residual of the candidate parameter model is generated according to the verification result, and the candidate parameter model with the smallest cumulative residual is determined as the representative model of the feature parameter group.

[0093] In practical applications, the verification result meets the preset verification conditions and includes the following two processes:

[0094] For each verification sample in the global verification set, the risk type represented by the verification result is consistent with the pre-labeled risk type, and the risk type includes risky and risk-free; and

[0095] The error between the risk prediction score represented by the verification result and the pre-marked risk standard score is within a specified interval.

[0096] For example, in the present application, risk levels may be roughly classified in advance, and all risk levels may be divided into risky and risk-free, wherein risky may correspond to levels above level 3 (including level 3) in Table 1, and risk-free may correspond to levels below level 2 (including level 2) in Table 1, that is, risky may correspond to HPTB defined as being in an unhealthy state, and risk-free may correspond to HPTB defined as being in a healthy state.

[0097] The first requirement: Under such a coarse classification, if the prediction score output by the candidate parameter model for all risky samples in the global validation set cannot be lower than 0.7 points (the threshold value of level 3), all risky samples are correctly predicted as risky samples. Similarly, risk-free samples also need to be correctly predicted as risk-free samples.

[0098] The second requirement: When the first requirement is met, if the difference between the predicted risk score and the standard risk score is too large, it means that the accuracy of the candidate parameter model is not high enough. In view of this, the error interval of the predicted risk score can be set. Only within this error interval can it be adopted as the final representative model. Specifically, according to the interval setting, the risk score predicted by the candidate parameter model should not fluctuate more than one level from the standard risk score. For example, if the actual risk level is 6, the predicted risk score range of 5-7 is acceptable. However, the premise is that such fluctuations need to meet the first requirement. For example, if the actual risk level is 3, it is unacceptable if the predicted risk level is 2 (because the risky samples are mistakenly predicted as risk-free samples).

[0099] When the first two conditions are met, the cumulative residual of each candidate parameter model can be calculated: (labelPoint-predictPoint), where n is the total number of validation samples in the global validation set, labelPoint is the true risk score, and predictPoint is the risk score predicted by the candidate parameter model. Finally, the candidate parameter model with the lowest cumulative residual can be selected as the representative model.

[0100] The technical solution provided in this embodiment uses a global validation set to verify the parameter model in the model cache pool, and can screen out a representative model that meets the requirements. The representative model can accurately predict the risk type of the validation sample. At the same time, for more detailed risk prediction scores, it can also meet the error requirements, thereby achieving a higher prediction accuracy.

[0101] In one embodiment, the method further includes: if the score representing the risk prediction level of the target engine exceeds a preset score threshold within a specified time period, disassembling the high-pressure turbine blades in the target engine and obtaining parameter data of the disassembled high-pressure turbine blades; and adding the obtained parameter data to the global verification set and the training data set respectively.

[0102] The technical solution provided in this embodiment is that if the risk prediction score of a target engine is high, there will be a risk of in-flight shutdown. At this time, the target engine can be disassembled and the parameter data of the high-pressure turbine blades obtained from the disassembly can be fed back to the HPTB's full life cycle database, thereby providing data support for the training data set and the global validation set, which is conducive to the subsequent training of a more accurate parameter model.

[0103] In this embodiment, after obtaining the representative models of each characteristic parameter group, these representative models can be used to perform multi-dimensional prediction on the actual data of the target engine to obtain multiple single-dimensional prediction results. Of course, in the prediction process, each representative model only predicts the factor parameters contained in the characteristic parameter group to which it belongs. For example, the representative model corresponding to the N2 speed will only predict the N2 speed data of the target engine.

[0104] S5: Based on the weight coefficients assigned to the respective characteristic parameter groups, weighted summation is performed on the multiple single-dimensional prediction results to generate a risk prediction level for the target engine, wherein the risk prediction level is used to characterize the service life of the high-pressure turbine blades in the target engine.

[0105] In this embodiment, after obtaining multiple single-dimensional prediction results, the weight coefficients of each characteristic parameter group predetermined in the above steps can be used to perform weighted summation on these single-dimensional prediction results to obtain a final risk score. The final risk score can be queried to obtain the corresponding risk prediction level of the target engine through Table 1. Each risk prediction level corresponds to the state of the HPTB, and the service life of the HPTB can be inferred.

[0106] The technical solution provided in this embodiment can decompose the multidimensional factor parameters related to the aircraft engine into multiple characteristic parameter groups. By training the model of each characteristic parameter group, the parameter model of each characteristic parameter group can be obtained. During the training process, according to the number of factor parameters contained in the characteristic parameter group, parameter models with different structures can be selected to make the feature processing process more accurate. Subsequently, by screening out the representative models of each characteristic parameter group, different single-dimensional prediction results can be predicted using the representative models. By weighted summing these single-dimensional prediction results, an accurate risk prediction level can be obtained. The risk prediction level can be used to characterize the service life of high-pressure turbine blades. Through the technical solution provided in this embodiment, the risk level of the engine can be accurately predicted without disassembling the engine, and the service life of the high-pressure turbine blades can be known, effectively avoiding the situation of air shutdown.

[0107] In one embodiment, if the score representing the risk prediction level of the target engine exceeds a preset score threshold (the preset score threshold can be flexibly set according to the actual application scenario) within a specified time period, the high-pressure turbine blades in the target engine can be disassembled, and parameter data of the disassembled high-pressure turbine blades can be obtained. The parameter data can be added to the global verification set and the training data set respectively.

[0108] The technical solution provided in this embodiment is that if the risk prediction score of a target engine is high, there will be a risk of in-flight shutdown. At this time, the target engine can be disassembled and the parameter data of the high-pressure turbine blades obtained from the disassembly can be fed back to the HPTB's full life cycle database, thereby providing data support for the training data set and the global validation set, which is conducive to the subsequent training of a more accurate parameter model.

[0109] See also Figure 4 One embodiment of the present application further provides a life prediction device for a high-pressure turbine blade in an aircraft engine, the device comprising:

[0110] A feature decomposition unit, used for counting multidimensional factor parameters related to the aircraft engine, and decomposing the multidimensional factor parameters into a plurality of feature parameter groups, wherein the multidimensional factor parameters are used to characterize engine performance parameters and / or flight status parameters, and the feature parameter group includes one or more factor parameters;

[0111] A model adaptation unit, for obtaining a data sample set of any characteristic parameter group, and selecting a parameter model adapted to the characteristic parameter group according to the number of factor parameters included in the characteristic parameter group; wherein, if the number of factor parameters included in the characteristic parameter group is more than one, the selected parameter model at least includes a parallel network branch for performing weight splitting;

[0112] A model training unit, used to train and verify the parameter model using the data sample set, and add the parameter model whose verification result meets the preset conditions to the model cache pool;

[0113] A prediction unit, used to select respective representative models for respective characteristic parameter groups from the model cache pool, and use the selected representative models to perform multi-dimensional prediction on the actual data of the target engine to obtain multiple single-dimensional prediction results;

[0114] A risk assessment unit is used to perform weighted summation on the multiple single-dimensional prediction results based on the weight coefficients assigned to each of the characteristic parameter groups to generate a risk prediction level for the target engine, wherein the risk prediction level is used to characterize the service life of the high-pressure turbine blades in the target engine.

[0115] The specific implementation of each unit module can be found in the description of the aforementioned method implementation method, which will not be repeated here.

[0116] See also Figure 5 The present application also provides a computer device, which includes a memory and a processor, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the above-mentioned method for predicting the life of high-pressure turbine blades in an aircraft engine is implemented.

[0117] like Figure 5 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 5 A processor 10 is taken as an example.

[0118] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0119] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0120] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0121] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0122] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0123] The embodiment of the present application also provides a computer-readable storage medium. The above method according to the embodiment of the present application can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0124] The embodiment of the present application provides a computer program product, which includes computer instructions, which are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method of any embodiment of the present application.

[0125] Although the embodiments of the present application are described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present application, and such modifications and variations are all within the scope defined by the appended claims.

[0126] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0127] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices, and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0128] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0129] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0130] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0131] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0132] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

[0133] Although the embodiments of the present application have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present application, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A life prediction method for high-pressure turbine blades in an aircraft engine, characterized in that: The method comprises: Counting multidimensional factor parameters related to the aircraft engine, and decomposing the multidimensional factor parameters into a plurality of characteristic parameter groups, wherein the multidimensional factor parameters are used to characterize engine performance parameters and / or flight status parameters, and the characteristic parameter groups include one or more factor parameters; For any characteristic parameter group, a data sample set of the characteristic parameter group is obtained, and a parameter model adapted to the characteristic parameter group is selected according to the number of factor parameters included in the characteristic parameter group; wherein, if the number of factor parameters included in the characteristic parameter group is more than one, the selected parameter model at least includes a parallel network branch for performing weight splitting; The parameter model is trained and verified using the data sample set, and the parameter model whose verification result meets the preset conditions is added to the model cache pool; Selecting respective representative models for respective characteristic parameter groups from the model cache pool, and using the selected representative models to perform multi-dimensional prediction on the actual data of the target engine, respectively, to obtain multiple single-dimensional prediction results; Based on the weight coefficients assigned to the respective characteristic parameter groups, the plurality of single-dimensional prediction results are weighted and summed to generate a risk prediction level for the target engine, wherein the risk prediction level is used to characterize the service life of the high-pressure turbine blades in the target engine.

2. The method according to claim 1, characterized in that Using the data sample set to train and verify the parameter model, and adding the parameter model whose verification result meets the preset conditions to the model cache pool includes: Extracting a first proportion of data from the data sample set as a global validation set, wherein the global validation set includes data samples representing a healthy state and data samples representing an unhealthy state; Performing designated batches of training for the parameter model, wherein each batch of training includes a designated number of iterative trainings, and each group of iterative training includes a designated number of rounds of training; wherein, for any batch of training, a second proportion of data is selected from a data sample set other than the global validation set as a local validation data set, and in any group of iterative training, a third proportion of data is selected as a test data set, and a fourth proportion of data is selected as a training data set; In each round of training, the parameter model is trained using the training data set, and after the training is completed, the trained parameter model is tested using the test data set to determine the gradient descent direction; After completing all rounds of training and testing in a set of iterative training, reselect the test data set and the training data set, and enter the next set of iterative training process based on the reselected test data set and the training data set; After completing the iterative training of all groups in a batch, the trained parameter model is verified using the local verification data set, and the parameter model with verification accuracy greater than or equal to the specified accuracy threshold is added to the model cache pool, and enters the training process of the next batch.

3. The method according to claim 2, characterized in that The method further comprises: If the verification accuracy is less than the specified accuracy threshold, the parameter model obtained by the current batch training will be discarded and the training process for the next batch will begin.

4. The method according to claim 2, characterized in that: The first ratio is 5%, the second ratio is 5%, the third ratio is 10%, the fourth ratio is 80%, the designated batch is 100 times, the designated number of groups is 15 groups, and the designated number of rounds is 30 rounds.

5. The method according to claim 1 or 2, characterized in that: The weight coefficients of each of the characteristic parameter groups are determined in the following manner: For any feature parameter group, identify the number of parameter models of the feature parameter group in the model cache pool, and determine the training batches of the parameter models of the feature parameter group during the training process; Calculating a ratio between the number of the parameter models and the training batches, and using the ratio as a model fitting parameter of the feature parameter group; The proportion of the model fitting parameters of the characteristic parameter group in the total model fitting parameters of each characteristic parameter group is calculated, and the proportion is used as the weight coefficient of the characteristic parameter group.

6. The method according to claim 1, characterized in that In the process of training and verifying the parameter model using the data sample set, the parameter model processes the data samples in the data sample set in the following manner: If the number of factor parameters included in the feature parameter group is 1, the data sample is normalized, and a three-layer fully connected network is used to perform feature classification on the normalized features, wherein an L2 regularization mechanism is added to each layer of the fully connected network for correction, and an early stopping mechanism is added to prevent overfitting, and a loss function is customized based on the ReLU function, and the loss function adds a penalty for the situation where the actual sample is in an unhealthy state but the predicted result is in a healthy state; If the number of factor parameters included in the feature parameter group is more than one, the weights of the parameter model are split, wherein the first part of the weights after the split enters the random forgetting branch, and the second part of the weights enters the feature processing branch, wherein the feature processing branch includes a convolution layer, a batch normalization layer, and a pooling layer; after the weights output by the random forgetting branch and the feature processing branch are integrated, they enter the attention layer for dynamic weight allocation; a three-layer fully connected network is used to classify the features after the dynamic weight allocation, wherein an L2 regularization mechanism is added to each layer of the fully connected network for correction, and an early stopping mechanism is added to prevent overfitting, and a loss function is customized based on the ReLU function, and the loss function adds a penalty for the situation where the actual sample is in an unhealthy state but the predicted result is in a healthy state.

7. The method according to claim 1, characterized in that Selecting respective representative models for each feature parameter group from the model cache pool includes: For any characteristic parameter group, a parameter model corresponding to the characteristic parameter group in the model cache pool is verified using a preset global verification set. If the parameter model correctly predicts the verification data representing the idle parking state in the global verification set, the parameter model is determined as a candidate parameter model; Obtain the verification result of the candidate parameter model for the global verification set, and when the verification result meets the preset verification condition, generate the cumulative residual of the candidate parameter model according to the verification result, and determine the candidate parameter model with the smallest cumulative residual as the representative model of the feature parameter group.

8. The method according to claim 7, characterized in that The verification result meets the preset verification condition including: For each verification sample in the global verification set, the risk type represented by the verification result is consistent with the pre-labeled risk type, and the risk type includes risky and risk-free; and The error between the risk prediction score represented by the verification result and the pre-marked risk standard score is within a specified interval.

9. The method according to claim 1, characterized in that: The method further comprises: If the score represented by the risk prediction level of the target engine exceeds a preset score threshold within a specified period of time, disassembling the high-pressure turbine blades in the target engine and obtaining parameter data of the disassembled high-pressure turbine blades; The obtained parameter data are added to the global verification set and the training data set respectively.

10. A life prediction device for high-pressure turbine blades in an aircraft engine, characterized in that: The device comprises: A feature decomposition unit, used for counting multidimensional factor parameters related to the aircraft engine, and decomposing the multidimensional factor parameters into a plurality of feature parameter groups, wherein the multidimensional factor parameters are used to characterize engine performance parameters and / or flight status parameters, and the feature parameter group includes one or more factor parameters; A model adaptation unit, for obtaining a data sample set of any characteristic parameter group, and selecting a parameter model adapted to the characteristic parameter group according to the number of factor parameters included in the characteristic parameter group; wherein, if the number of factor parameters included in the characteristic parameter group is more than one, the selected parameter model at least includes a parallel network branch for performing weight splitting; A model training unit, used to train and verify the parameter model using the data sample set, and add the parameter model whose verification result meets the preset conditions to the model cache pool; A prediction unit, used to select respective representative models for respective characteristic parameter groups from the model cache pool, and use the selected representative models to perform multi-dimensional prediction on the actual data of the target engine to obtain multiple single-dimensional prediction results; A risk assessment unit is used to perform weighted summation on the multiple single-dimensional prediction results based on the weight coefficients assigned to each of the characteristic parameter groups to generate a risk prediction level for the target engine, wherein the risk prediction level is used to characterize the service life of the high-pressure turbine blades in the target engine.

11. A computer device, characterized in that: The computer device comprises a memory and a processor, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the method according to any one of claims 1 to 9 is implemented.

Images