A method for predicting aircraft engine failure

Abstract

The present invention relates to an aircraft engine fault prediction method, which belongs to the technical field of equipment fault early warning, and solves the problem of inaccurate fault prediction caused by individual differences in operating environment, service life and other aspects of different aircraft engines in the prior art. The method comprises: obtaining full life cycle data of multiple training aircraft engines; the full life cycle data is divided into normal data and abnormal data; the abnormal data contains fault early warning data; using the normal data in each training aircraft engine to train the SOM model of the corresponding aircraft engine, and obtain the MQE of the fault early warning data; based on the average value of the MQE of the fault early warning data in multiple training aircraft engines, determining the MQE range corresponding to the early warning interval; using multiple test aircraft engines to test the accuracy of the MQE range respectively, if the proportion of the test aircraft engines whose accuracy exceeds the accuracy threshold exceeds the first test pass rate threshold, then based on the MQE range, the fault prediction of the aircraft engine to be predicted is performed.

Description

Technical Field

[0001] The invention relates to the technical field of equipment failure early warning, and in particular to a method for predicting aircraft engine failure. Background Art

[0002] In actual industrial production, the lifespan of a small standard part or a large complex system such as an engine is limited, and maintenance (or replacement) is required after a period of operation. Therefore, timely warning of equipment failure can effectively reduce the occurrence of accidents. At present, the fault prediction technology of mechanical equipment mainly focuses on signal analysis and intelligent prediction. In terms of signal analysis, relevant practitioners need to have a good knowledge base and experience; in terms of intelligent prediction, the application of machine learning related algorithms has achieved remarkable results, mainly including support vector machines (SVM), artificial neural networks (ANN), kernel methods, and convolutional neural networks. Pattern recognition methods such as these methods need to fit model parameters through a large amount of historical abnormal data during the model training process. However, for actual industrial scenarios, the normal state data of the equipment can be easily and accurately obtained, while the fault samples are often difficult to collect, which poses a great challenge to the common supervised learning model.

[0003] The self-organizing map network (hereinafter referred to as SOM) is an unsupervised artificial neural network that simulates the characteristics of signal processing in the human brain and gradually optimizes the network by relying on competition between neurons. The SOM algorithm has a simple structure and a small number of parameters, so it is widely used in speech recognition, clustering, and outlier removal. In the prior art, there have been studies that combine the SOM algorithm with the immune neural network to predict and predict the faults of the fuel system in aircraft engines. This method uses the BP neural network for fault classification based on the use of the SOM algorithm for data processing, and still requires sufficient abnormal data samples. However, this study did not take into account the individual differences in the operating environment, service life, etc. of different aircraft engines, resulting in low accuracy of fault prediction results. Summary of the invention

[0004] In view of the above analysis, an embodiment of the present invention aims to provide an aircraft engine fault prediction method to solve the problem in the prior art of low accuracy of fault prediction results caused by individual differences in operating environment, service life, etc. of different aircraft engines.

[0005] A method for predicting aircraft engine failure, comprising:

[0006] Acquire full life cycle data of multiple training aircraft engines; the full life cycle data is divided into normal data and abnormal data according to the length of use; the abnormal data includes fault warning data within the warning interval in the full life cycle;

[0007] Using the normal data in each of the training aircraft engines, training the SOM model of the corresponding aircraft engine, and obtaining the minimum quantization error MQE of the fault warning data in the aircraft engine;

[0008] Determine a minimum quantization error MQE range corresponding to the warning interval based on an average value of minimum quantization errors MQE of the fault warning data in the engines of the plurality of training aircraft;

[0009] The warning accuracy of the minimum quantization error MQE range is tested respectively using multiple test aircraft engines, and the proportion of test aircraft engines whose accuracy exceeds the accuracy threshold is taken as the test pass rate. If the test pass rate exceeds the first test pass rate threshold, fault prediction of the aircraft engine to be predicted is performed based on the minimum quantization error MQE range.

[0010] Based on the above scheme, this scheme also makes the following improvements:

[0011] Based on the further improvement of the above method, each test aircraft engine tests the accuracy of the minimum quantization error MQE range by performing the following steps:

[0012] Using normal data in the test aircraft engine to train the SOM model of the test aircraft engine, and obtaining the minimum quantization error MQE of the abnormal data in the test aircraft engine;

[0013] Obtaining an abnormal data interval of a minimum quantization error MQE of the abnormal data within a range of the minimum quantization error MQE;

[0014] The accuracy is obtained based on the ratio of the intersection of the abnormal data interval and the warning interval to the length of the warning interval.

[0015] A further improvement based on the above method further includes: if the test pass rate is between a second test pass rate threshold and the first test pass rate threshold, adjusting the SOM model parameters of the test aircraft engine whose accuracy is lower than the accuracy threshold or the proportion of normal data in the test aircraft engine, and retraining the SOM model of the test aircraft engine, and reacquiring the accuracy and test pass rate;

[0016] Among them, the second test pass rate threshold is lower than the first test pass rate threshold.

[0017] A further improvement based on the above method also includes: if the test pass rate is lower than the second test pass rate threshold, adjusting one or more of the SOM model parameters of the training aircraft engine, the proportion of normal data in the full life cycle data, and the range of the warning interval, and redetermining the minimum quantization error MQE range, and re-obtaining the accuracy rate and test pass rate.

[0018] Based on the further improvement of the above method, the parameters of the SOM model for training aircraft engines and testing aircraft engines are adjusted, including adjusting one or more parameters among the learning rate initialization parameters, the initial winning neighborhood, and the initial value of the neuron weight in the SOM model.

[0019] Based on further improvement of the above method, the accuracy threshold is 70%.

[0020] Based on further improvement of the above method, the first test pass rate is 90%, and the second test pass rate is 80%.

[0021] Based on the further improvement of the above method, the fault prediction of the aircraft engine to be predicted based on the minimum quantization error MQE range includes:

[0022] Each time normal data of an aircraft engine to be predicted is collected, the SOM model of the aircraft engine to be predicted is trained using the collected normal data of the aircraft engine to be predicted;

[0023] Each time abnormal data of an aircraft engine to be predicted is collected, a minimum quantization error MQE of the abnormal data in the aircraft engine to be predicted is obtained;

[0024] If the minimum quantization error MQE of the abnormal data in the aircraft engine to be predicted is within the minimum quantization error MQE range, fault prediction of the aircraft engine to be predicted is performed.

[0025] Based on further improvement of the above method, the lower limit of the minimum quantization error MQE range is 80% of the average value of the minimum quantization error MQE, and the upper limit is 120% of the average value of the minimum quantization error MQE.

[0026] Based on the further improvement of the above method, the method of using the normal data in each aircraft engine to train the SOM model of the current aircraft engine and obtain the minimum quantization error MQE of the fault warning data in the current aircraft engine includes:

[0027] The normal data in each aircraft engine is used to train the SOM model of the current aircraft engine to obtain the best matching unit;

[0028] The minimum value of the distance between the fault warning data and the best matching unit is used as the minimum quantization error MQE of the fault warning data.

[0029] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0030] First, the aircraft engine fault prediction method provided by the present invention fully considers the individual differences in the operating environment, service life, factory settings and wear degree of different aircraft engines, and trains a plurality of SOM models of training aircraft engines respectively to obtain accurate SOM model training results and corresponding minimum quantization error MQE corresponding to the aircraft engine;

[0031] Second, the aircraft engine fault prediction method provided by the present invention takes into account that the timings of fault warnings of different aircraft engines of the same type are roughly the same (i.e., the warning interval corresponding to each training aircraft engine). Therefore, the minimum quantization error MQE range corresponding to the warning interval can be determined by calculating the average value of the minimum quantization error MQE of fault warning data in multiple training aircraft engines. The minimum quantization error MQE range determined in this way can maximize the integration of individual differences of different aircraft engines and can be used for fault warnings of the same type of aircraft engines.

[0032] Thirdly, the aircraft engine fault prediction method provided by the present invention predicts the fault of the aircraft engine to be predicted based on the minimum quantization error MQE range, and can timely remind the user to repair or replace the aircraft engine, effectively reducing the number of accidental maintenance of the aircraft engine and ensuring the safe use of the aircraft engine;

[0033] Fourth, the aircraft engine fault prediction method provided by the present invention also uses the accuracy of the minimum quantization error MQE range tested by multiple test aircraft engines. When the test pass rate exceeds the first test pass rate threshold, the fault prediction of the aircraft engine to be predicted is performed based on the minimum quantization error MQE range, thereby ensuring that the applicability of the determined minimum quantization error MQE range is strong enough.

[0034] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like components throughout the drawings.

[0036] Figure 1 This is a flow chart of the aircraft engine fault prediction method provided in Embodiment 1 of the present invention;

[0037] Figure 2 It is the MQE distribution diagram of the fault warning data provided in Example 2 of the present invention;

[0038] Figure 3 This is a MQE value variation trend diagram of the measured data of the No. 1 test device provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0039] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0040] Example 1

[0041] Embodiment 1 of the present invention discloses a method for predicting aircraft engine failures, the flow chart of which is as follows: Figure 1 As shown, the following steps are included:

[0042] Step S1: acquiring full life cycle data of multiple training aircraft engines; the full life cycle data is divided into normal data and abnormal data according to the length of use; the abnormal data includes fault warning data within the warning interval in the full life cycle;

[0043] Step S2: using the normal data in each of the training aircraft engines to train the SOM model of the corresponding aircraft engine, and obtaining the minimum quantization error MQE of the fault warning data in the aircraft engine;

[0044] Step S3: determining a minimum quantization error MQE range corresponding to the warning interval based on an average value of minimum quantization errors MQE of the fault warning data in the engines of the plurality of training aircraft;

[0045] Step S4: Use multiple test aircraft engines to test the warning accuracy of the minimum quantization error MQE range respectively, and take the proportion of test aircraft engines whose accuracy exceeds the accuracy threshold as the test pass rate. If the test pass rate exceeds the first test pass rate threshold, then based on the minimum quantization error MQE range, fault prediction of the aircraft engine to be predicted is performed.

[0046] Compared with the prior art, the aircraft engine fault prediction method provided by the embodiment of the present invention has the following advantages: firstly, it fully considers the individual differences in the operating environment, service life, factory settings and wear degree of different aircraft engines, and trains the SOM models of multiple training aircraft engines respectively to obtain accurate SOM model training results and corresponding minimum quantization error MQE of the corresponding aircraft engines; secondly, considering that the timing of fault warning of different aircraft engines of the same type is roughly the same (that is, the warning interval corresponding to each training aircraft engine), therefore, the minimum quantization error MQE range corresponding to the warning interval can be determined by calculating the average value of the minimum quantization error MQE of the fault warning data in multiple training aircraft engines. The minimum quantization error MQE range determined by the method can maximize the integration of individual differences of different aircraft engines and can be used for fault warning of the same type of aircraft engines; furthermore, based on the minimum quantization error MQE range, fault prediction of the aircraft engine to be predicted can timely remind the user to repair or replace the aircraft engine, effectively reduce the number of unexpected maintenance times of the aircraft engine, and ensure the safe use of the aircraft engine; finally, the method also uses multiple test aircraft engines to test the accuracy of the minimum quantization error MQE range. When the test pass rate exceeds the first test pass rate threshold, the fault prediction of the aircraft engine to be predicted is performed based on the minimum quantization error MQE range, thereby ensuring that the applicability of the determined minimum quantization error MQE range is strong enough.

[0047] It should be noted that the training aircraft engine, the aircraft engine to be predicted and the test aircraft engine introduced later in this embodiment are all aircraft engines of the same type. At the same time, the training aircraft engine and the test aircraft engine both contain the full life cycle data of the aircraft engine from the healthy state, the decay state to the fault state.

[0048] Preferably, in step S4, in order to avoid the low accuracy of the minimum quantization error MQE range caused by the individual differences of a single test aircraft engine, in this embodiment, the warning accuracy corresponding to multiple test aircraft engines is calculated at the same time, and the test pass rate is obtained based on the warning accuracy rate; if the test pass rate exceeds the first test pass rate threshold, it indicates that the determined minimum quantization error MQE range has a high accuracy and can be directly used for fault prediction of the aircraft engine to be predicted. Among them, the first test pass rate is preferably set to 90% (this parameter can be adjusted according to the test accuracy requirements).

[0049] Preferably, in step S4, each test aircraft engine tests the accuracy of the minimum quantization error MQE range by executing the following steps:

[0050] Using normal data in the test aircraft engine to train the SOM model of the test aircraft engine, and obtaining the minimum quantization error MQE of the abnormal data in the test aircraft engine;

[0051] Obtaining an abnormal data interval of a minimum quantization error MQE of the abnormal data within a range of the minimum quantization error MQE;

[0052] The accuracy rate is obtained based on the ratio of the intersection of the abnormal data interval and the warning interval to the duration of the warning interval; it should be noted that the ratio of the intersection of the abnormal data interval and the warning interval (i.e., the warning interval determined according to the test aircraft engine) to the duration of the warning interval represents the degree of overlap between the warning result corresponding to the test aircraft engine and the warning interval set during the training process, and can be used to characterize the accuracy rate.

[0053] If the accuracy exceeds the set accuracy threshold, the current test aircraft engine meets the accuracy requirement, indicating that the determined minimum quantization error MQE range can meet the fault prediction requirements of the current test aircraft engine.

[0054] Preferably, the accuracy threshold can be determined according to the test accuracy requirement. For example, in the absence of clear fault label data, the accuracy threshold can be set to 70% according to the general requirements of the industry.

[0055] In this embodiment, if the test pass rate does not exceed the first test pass rate threshold, the reasons for this situation can be analyzed from the following two aspects:

[0056] (1) Some issues related to testing aircraft engines:

[0057] Since this situation is caused by improper parameter selection of some test aircraft engines during the test process, and the test processes of different test aircraft engines do not affect each other, the test pass rate caused by improper parameter selection of individual test aircraft engines is generally not too low, and illustratively, it will not be lower than the second test pass rate threshold (preferably set to 80%); when the test pass rate is between the second test pass rate threshold and the first test pass rate threshold, the relevant data of the current test aircraft engine is adjusted preferentially, and the correctness is recalculated, specifically, including the following situations:

[0058] 1) The SOM model parameters of some test aircraft engines were improperly set:

[0059] adjusting the SOM model parameters of the test aircraft engine, retraining the SOM model of the test aircraft engine, and reacquiring the accuracy; if the reacquired warning accuracy exceeds the accuracy threshold, it indicates that the low prediction accuracy may be caused by improper setting of the SOM model parameters;

[0060] 2) The proportion of normal data of some test aircraft engines was improperly selected:

[0061] Adjust the proportion of normal data in the test aircraft engine, retrain the SOM model of the test aircraft engine, and re-acquire the warning accuracy. If the re-acquired warning accuracy exceeds the accuracy threshold, it indicates that the low prediction accuracy may be caused by improper selection of normal data for the test aircraft engine;

[0062] After recalculating all test aircraft engines whose accuracy is lower than the accuracy threshold, recalculate the test pass rate; if the recalculated test pass rate exceeds the first test pass rate threshold, then based on the minimum quantization error MQE range, perform fault prediction of the aircraft engine to be predicted; if the recalculated test pass rate is between the second test pass rate and the first test pass rate, then adjust the above parameters of the test aircraft engine whose accuracy is lower than the accuracy threshold again, and try to adjust the SOM model parameters of the aircraft engine and the proportion of normal data of the test aircraft engine at the same time, and calculate the warning accuracy again. If the test pass rate after repeated tests is still lower than the first test pass rate, or the test pass rate of a certain test is lower than the second test pass rate threshold, consider whether there is a problem with the determined minimum quantization error MQE range. At this time, adjust one or more of the SOM model parameters of the training aircraft engine, the proportion of normal data in the full life cycle data, and the range of the warning interval, and re-determine the minimum quantization error MQE range and re-acquire the warning accuracy.

[0063] The above process of testing and re-determining the minimum quantization error MQE range is relatively complicated and may require repeated experiments to obtain the optimal minimum quantization error MQE range.

[0064] It should be noted that adjusting the SOM model parameters for training aircraft engines and testing aircraft engines includes adjusting one or more parameters among the learning rate initialization parameters, the initial winning neighborhood, and the initial value of the neuron weight in the SOM model.

[0065] Preferably, step S2 comprises:

[0066] Step S21: using the normal data in each training aircraft engine to train the SOM model of the current aircraft engine to obtain the best matching unit;

[0067] It should be noted that SOM is a neural network algorithm that can map high-dimensional data to low-dimensional data. In the training phase, the distance between the features of each input sample and the neurons in the mapping layer is calculated, and the neuron closest to the input sample in the mapping layer is found, and this neuron is defined as the best matching unit (BMU).

[0068] Step S22: taking the minimum value of the distance between the fault warning data and the best matching unit as the minimum quantization error MQE of the fault warning data. The distance between the fault warning data and the best matching unit can be obtained by a Euclidean distance calculation method.

[0069] Preferably, step S4 further predicts the failure of the aircraft engine by executing the following steps:

[0070] Step S41: collecting normal data of each aircraft engine to be predicted, and using the collected normal data of the aircraft engine to be predicted to train the SOM model of the aircraft engine to be predicted;

[0071] Step S42: each time abnormal data of an aircraft engine to be predicted is collected, a minimum quantization error MQE of the abnormal data in the aircraft engine to be predicted is obtained;

[0072] Step S43: If the minimum quantization error MQE of the abnormal data in the aircraft engine to be predicted is within the minimum quantization error MQE range, then a fault prediction of the aircraft engine to be predicted is performed.

[0073] Preferably, a lower limit of the range of the minimum quantization error MQE is 80% of an average value of the minimum quantization error MQE, and an upper limit of the range of the minimum quantization error MQE is 120% of the average value of the minimum quantization error MQE.

[0074] During the implementation of this embodiment, normal data, abnormal data and warning intervals can be determined in the following manner: normal data refers to the data of the first 60% of the usage time in the full life cycle data of the aircraft engine; abnormal data refers to the data of the last 40% of the usage time in the full life cycle data of the aircraft engine; the warning interval refers to the range of 88%-92% of the usage time in the full life cycle data of the aircraft engine.

[0075] Preferably, the full life cycle data of the aircraft engine includes: the full life cycle data of the aircraft engine includes: engine intake temperature, exhaust temperature, lubricating oil temperature, lubricating oil pressure, compressed air pressure and vibration speed parameters, etc.

[0076] Example 2

[0077] Embodiment 2 of the present invention discloses the training and testing process of the aircraft engine fault early warning method based on SOM-MQE in Embodiment 1, which is specifically described as follows:

[0078] The full life cycle data of 218 aircraft engines of the same type are selected. The full life cycle data of each aircraft engine includes the entire process of the aircraft engine from a healthy state, a decaying state to a faulty (scrapped) state. The specific content included in the full life cycle data can be set with reference to Example 1. Assuming that the rated service life of such aircraft engines is 357 cycles, the duration corresponding to the full life cycle of different aircraft engines may not be different due to differences in factory delivery and different degrees of wear.

[0079] During the implementation process, the training set and the test set are first divided, and the full life cycle data of the first 100 aircraft engines are selected as the training set (these 100 aircraft engines are all used as training aircraft engines), and the full life cycle data of the last 118 aircraft engines are selected as the test set (these 118 aircraft engines are all used as test aircraft engines). And the normal data, abnormal data and warning interval are divided according to the method in Example 1.

[0080] By executing steps S2 and S3 in Example 1 (before executing the steps, a step of normalizing the whole life cycle data may be added), an MQE distribution diagram of the fault warning data is obtained, such as Figure 2 The average value of the minimum quantization error MQE of the data in the warning interval (i.e., fault warning data) in all training sets is 2.64, which is multiplied by 0.8 to obtain the lower limit of the minimum quantization error MQE range (i.e. Figure 2 The lower limit of the MQE threshold of the fault warning data in the middle) 2.11, multiplied by 1.2 to get the upper limit of the minimum quantization error MQE range (i.e. Figure 2 The upper limit of the MQE threshold of the fault warning data is 3.16.

[0081] After the minimum quantization error MQE range is determined, the accuracy of the minimum quantization error MQE range needs to be tested. Similarly, the full life cycle data in the test set is normalized first, and then the first 60% of the use cycle of the test aircraft engine in the test set is used as normal data to train the SOM model. After the model training is completed, the remaining data of each engine in the test set is used to calculate its MQE value with the trained SOM model. Then the warning accuracy is calculated according to the corresponding content in Example 1.

[0082] Taking an engine in the test set as an example (called test equipment No. 1), the first 60% cycle data of this aircraft engine is selected as normal data to initialize the SOM model, and the remaining cycle data is used for verification, such as Figure 3 As shown, the broken line is the MQE value of the remaining cycle data of this aircraft engine deviating from the SOM model. Figure 3It can be clearly seen that as the operating time of the aircraft engine increases, the MQE value continues to increase, gradually deviating from the normal MQE model, and the risk of aircraft engine failure is increasing. Figure 3 The two horizontal lines in are the upper and lower limits of the minimum quantization error MQE range obtained by training the training set, that is, when the MQE value in the tested data is within this range, the equipment is considered to be in a fault warning state. The MQE value of the tested data in all test set data is calculated. According to the calculation method in Example 1, the warning accuracy of the warning model can be calculated to be 74.81%, which meets the 70% equipment fault warning accuracy requirement of the industry in the absence of clear fault label data.

[0083] In summary, the aircraft engine fault warning method based on SOM-MQE provided in this embodiment can make full use of the massive health data and usage experience in the historical data to build a benchmark model in the absence of a clear fault label, and judge the health status of the aircraft engine in real time based on the deviation between the observed data and the normal benchmark model (SOM model), so as to quickly and effectively provide fault warning for the aircraft engine, reduce the number of maintenance times caused by unexpected aircraft engine failures, and have significant reference significance for practical applications in the industry.

[0084] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0085] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for predicting aircraft engine failure, characterized in that: include: Obtain full life cycle data for multiple training aircraft engines; The full life cycle data is divided into normal data and abnormal data according to the usage duration; The abnormal data includes fault warning data within the warning interval in the entire life cycle; Using the normal data in each of the training aircraft engines, training the SOM model of the corresponding aircraft engine, and obtaining the minimum quantization error MQE of the fault warning data in the aircraft engine; implement: Using the normal data in each of the training aircraft engines, training the SOM model of the corresponding aircraft engine to obtain the best matching unit; The minimum value of the distance between the fault warning data and the best matching unit is used as the minimum quantization error MQE of the fault warning data; Determine a minimum quantization error MQE range corresponding to the warning interval based on an average value of minimum quantization errors MQE of the fault warning data in the engines of the plurality of training aircraft; Using a plurality of test aircraft engines to respectively test the warning accuracy of the minimum quantization error MQE range, taking the proportion of the test aircraft engines whose warning accuracy exceeds the accuracy threshold as the test pass rate, and if the test pass rate exceeds the first test pass rate threshold, performing fault prediction of the aircraft engine to be predicted based on the minimum quantization error MQE range; Each test aircraft engine tests the warning accuracy of the minimum quantization error MQE range by performing the following steps: Using normal data in the test aircraft engine to train the SOM model of the test aircraft engine, and obtaining the minimum quantization error MQE of the abnormal data in the test aircraft engine; Obtaining an abnormal data interval of a minimum quantization error MQE of the abnormal data within a range of the minimum quantization error MQE; The warning accuracy rate is obtained based on the ratio of the intersection of the abnormal data interval and the warning interval to the length of the warning interval; Among them, the training aircraft engine, the test aircraft engine and the aircraft engine to be predicted are all aircraft engines of the same type.

2. The aircraft engine failure prediction method according to claim 1, characterized in that: Also includes: If the test pass rate is between the second test pass rate threshold and the first test pass rate threshold, adjusting the SOM model parameters of the test aircraft engine whose warning accuracy is lower than the accuracy threshold or the proportion of normal data in the test aircraft engine, retraining the SOM model of the test aircraft engine, and reacquiring the warning accuracy and the test pass rate; Among them, the second test pass rate threshold is lower than the first test pass rate threshold.

3. The aircraft engine failure prediction method according to claim 2, characterized in that: Also includes: If the test pass rate is lower than the second test pass rate threshold, adjust one or more of the SOM model parameters of the training aircraft engine, the proportion of normal data in the full life cycle data, and the range of the warning interval, and redetermine the minimum quantization error MQE range, and re-obtain the warning accuracy rate and test pass rate.

4. The aircraft engine failure prediction method according to claim 3, characterized in that: Adjusting the parameters of the SOM model for training and testing aircraft engines includes adjusting one or more parameters among a learning rate initialization parameter, an initial winning neighborhood, and an initial value of a neuron weight in the SOM model.

5. The aircraft engine failure prediction method according to any one of claims 2 to 4, characterized in that: The accuracy threshold is 70%.

6. The aircraft engine failure prediction method according to claim 5, characterized in that: The first test pass rate threshold is 90%, and the second test pass rate threshold is 80%.

7. The aircraft engine failure prediction method according to claim 1, characterized in that: The method of performing fault prediction of the aircraft engine to be predicted based on the minimum quantization error MQE range includes: Each time normal data of an aircraft engine to be predicted is collected, the SOM model of the aircraft engine to be predicted is trained using the collected normal data of the aircraft engine to be predicted; Each time abnormal data of an aircraft engine to be predicted is collected, a minimum quantization error MQE of the abnormal data in the aircraft engine to be predicted is obtained; If the minimum quantization error MQE of the abnormal data in the aircraft engine to be predicted is within the minimum quantization error MQE range, fault prediction of the aircraft engine to be predicted is performed.

8. The aircraft engine failure prediction method according to claim 1, characterized in that: A lower limit of the minimum quantization error MQE range is 80% of an average value of the minimum quantization error MQE, and an upper limit is 120% of the average value of the minimum quantization error MQE.

Images