Aero-engine vibration fault early warning method based on working condition self-adaption

Abstract

The invention relates to an aero-engine vibration fault early warning method based on working condition self-adaption. The method comprises the steps that vibration signals and rotating speed data in the operation process of an aero-engine are synchronously collected through a vibration sensor and a rotating speed sensor; calculating an instantaneous rotating speed change rate based on the collected rotating speed data, and determining the working condition of the aero-engine according to the instantaneous rotating speed change rate; in the transient working condition freezing fault early warning mechanism, when the working condition of the aero-engine is a steady-state working condition, rotating speed data are discretized into N partitions, and vibration time sequence characteristics are predicted by adopting trend prediction models of the corresponding partitions to obtain predicted values; and calculating an absolute residual error between the characteristics of the collected vibration signal of the aero-engine and the predicted value, and carrying out double-window dynamic baseline management and graded early warning based on the absolute residual error. The aero-engine fault detection method can improve the aero-engine fault detection accuracy.

Description

Technical Field

[0001] This invention relates to the field of aero-engine health management technology, and in particular to an aero-engine vibration fault early warning method based on operating condition adaptation. Background Technology

[0002] As the core power component of an aircraft, the stability and reliability of its operation directly affect flight safety. During engine operation, vibration is a key indicator of its health; abnormal vibration often foreshadows potential malfunctions such as rotor imbalance, bearing wear, and blade failure. However, the operating conditions of aircraft engines are complex and varied, encompassing different phases such as takeoff, cruise, and landing, with significant differences in parameters such as engine speed and load across these phases.

[0003] Traditional early warning methods for aero-engine vibration faults often employ a fixed threshold method. This involves setting a fixed vibration threshold, and triggering an alert when the measured vibration value exceeds this threshold. However, this method has significant drawbacks. Under varying operating conditions, the vibration characteristics of the engine during normal operation can change considerably, making it difficult for the fixed threshold to accurately distinguish between normal and faulty vibrations, leading to false alarms or missed alarms. For example, during engine takeoff, the engine speed increases rapidly, causing significant fluctuations in the vibration signal. In such cases, the fixed threshold may cause false alarms. Conversely, in the early stages of minor faults, the vibration changes are small, and the fixed threshold may fail to detect the fault in time, severely impacting the effectiveness and reliability of the early warning system. Furthermore, existing model-based early warning methods often fail to adequately consider the differences in vibration characteristics under different engine speed conditions. The models lack adaptability and accuracy, failing to meet the ever-increasing safety requirements of aero-engines. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a vibration fault early warning method for aero-engines based on operating condition adaptation, which can improve the accuracy of aero-engine fault detection.

[0005] The technical solution adopted by this invention to solve its technical problem is: to provide a vibration fault early warning method for aero-engines based on operating condition adaptation, comprising the following steps:

[0006] Vibration signals and speed data of the aero-engine are collected synchronously during operation using vibration sensors and speed sensors.

[0007] The instantaneous speed change rate is calculated based on the collected speed data, and the operating condition of the aero-engine is determined based on the instantaneous speed change rate.

[0008] When the aero-engine is in steady-state operation, the speed data is discretized into N partitions, and the trend prediction model of the corresponding partition is used to predict the vibration time series characteristics to obtain the predicted value.

[0009] The absolute residual between the characteristics of the collected vibration signals of the aero-engine and the predicted values ​​is calculated, and baseline management and graded early warning are performed based on the absolute residual.

[0010] The determination of the aero-engine operating condition based on the instantaneous speed change rate is specifically as follows: when the absolute value of the difference between the current speed and the previous speed divided by the sampling interval is greater than or equal to a preset value, the aero-engine is in transient operating condition; when the absolute value of the difference between the current speed and the previous speed divided by the sampling interval is lower than a preset value, the aero-engine is in steady-state operating condition.

[0011] When discretizing the speed data into N partitions, non-uniform discretization is used. In the low speed range, partitions are made at intervals of 1000 RPM, and in the high speed range, partitions are made at intervals of 500 RPM. The low speed range and the high speed range are continuous intervals.

[0012] The trend prediction model for the corresponding partition takes the rotational speed data of the partition as input and the characteristics of the vibration signal of the aero-engine as output; the training data for the trend prediction model for the corresponding partition comes from the normal operation data of the corresponding partition for more than K consecutive hours.

[0013] The aforementioned air-engine vibration fault early warning method based on operating condition adaptation further includes, when the prediction error of the trend prediction model of the corresponding partition exceeds twice the historical average prediction error of the partition for M consecutive times, the retraining process of the trend prediction model of the corresponding partition is triggered. The training data for retraining includes the latest collected p-hour normal operation data and Kp-hour data randomly selected from the original training data.

[0014] The baseline management specifically involves updating the mean and standard deviation of the absolute residuals of the current partition based on a sliding window statistical method, and establishing a dynamic baseline based on the mean and standard deviation.

[0015] For short-term windows, the mean and standard deviation of the absolute residuals of the current partition are calculated using an exponentially weighted moving average algorithm.

[0016] For long-term windows, a fixed window length is used to record the historical maximum absolute residual. Each time the long-term window is updated, new data is added to the window while the oldest data is removed. When the absolute residual of the new data exceeds the historical maximum absolute residual by a preset multiple, a baseline update is triggered.

[0017] The tiered early warning system specifically implements a three-tiered early warning mechanism based on the degree of deviation between the absolute residual and the baseline, as follows:

[0018] When the absolute residual is greater than the sum of the mean and three times the standard deviation, a level one warning is triggered, an abnormal event is identified, and the abnormal event is continuously accumulated.

[0019] When the number of abnormal events accumulates to a preset number, a level-two warning is triggered, and maintenance personnel are alerted to pay attention to the operating status of the aircraft engine.

[0020] When the absolute residual is greater than the sum of the mean and 6 times the standard deviation, a level 3 warning is triggered, and protective actions are initiated.

[0021] When the aircraft engine is in a transient operating condition, the fault warning mechanism is frozen.

[0022] The technical solution adopted by the present invention to solve its technical problem is: to provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-mentioned air-engine vibration fault early warning method based on operating condition adaptation.

[0023] The technical solution adopted by the present invention to solve its technical problem is: to provide a computer-readable storage medium on which a computer program is stored, wherein when the computer program is executed by a processor, the steps of the above-mentioned air-engine vibration fault early warning method based on operating condition adaptation are implemented.

[0024] Beneficial effects

[0025] By adopting the above-mentioned technical solutions, this invention has the following advantages and positive effects compared with the prior art: This invention employs different processing strategies for different operating conditions. Partitioned modeling is performed under steady-state conditions, enabling the model to adapt to the differences in engine vibration characteristics at different speeds; the fault warning mechanism is frozen under transient conditions, effectively avoiding misjudgments during changes in operating conditions and greatly improving the adaptability of the warning method to complex operating conditions. Attached Figure Description

[0026] Figure 1 This is a flowchart of the first embodiment of the air-to-air engine vibration fault early warning method based on operating condition adaptation of the present invention;

[0027] Figure 2 This is a flowchart of the training process for the trend prediction model of the corresponding partition in the first embodiment of the present invention.

[0028] Figure 3 This is a schematic diagram of dynamic baseline management in the first embodiment of the present invention. Detailed Implementation

[0029] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0030] The first embodiment of this invention relates to a vibration fault early warning method for aero-engines based on adaptive operating conditions. This method, through accurate identification and adaptive processing of engine operating conditions, combined with advanced time-series prediction models and dynamic baseline management, solves the problem of high false alarm rates in traditional methods under varying operating conditions, improves the sensitivity and reliability of aero-engine vibration fault early warning, and ensures the safe and stable operation of aero-engines. Figure 1 As shown, the aircraft engine vibration fault early warning method of this embodiment includes the following steps:

[0031] Step S100: Data acquisition.

[0032] In this step, vibration signals and speed data during the operation of the aero-engine are simultaneously acquired using vibration sensors and speed sensors. To eliminate interference from high-frequency noise, this embodiment can also perform anti-aliasing filtering on the acquired raw data. This processing can be achieved using an FIR low-pass filter with a cutoff frequency half that of the sampling frequency.

[0033] Step S200: Operating condition determination.

[0034] In this step, the instantaneous speed change rate ΔRPM / Δt is calculated using the sliding difference method. This parameter is used to determine the current operating condition of the aero-engine. The formula for calculating the instantaneous speed change rate ΔRPM / Δt is as follows:

[0035]

[0036] Among them: RPM t-1 It is the rotational speed at time t-1, in RPM. t It is the rotational speed at time t, t s It is the sampling interval.

[0037] When ΔRPM / Δt≥100RPM / s, it indicates that the aero-engine is in a transient operating mode. This mode is short in duration, the aero-engine's operating state changes drastically, and the vibration characteristics are unstable. Therefore, the fault warning mechanism is frozen at this time.

[0038] When ΔRPM / Δt < 100RPM / s, it indicates that the aero-engine is in steady-state operating mode. At this time, the aero-engine operates relatively stably and the vibration characteristics have certain regularity, so the fault early warning mechanism is activated.

[0039] Step S300: Model prediction.

[0040] Under steady-state conditions, non-uniform discretization is used to discretize the speed data into N partitions. In the low speed range (0-5000 RPM), the partitions are divided at intervals of 1000 RPM, and in the high speed range (above 5000 RPM), the partitions are divided at intervals of 500 RPM.

[0041] like Figure 2 As shown, this implementation pre-trains a corresponding trend prediction model for each partition. This trend prediction model can be trained based on an LSTM network model, which can effectively process time series data and learn the temporal characteristics and variation patterns of vibration signals under different speed partitions. When the aero-engine is in a certain steady-state speed partition, the trend prediction model corresponding to that partition can be called to perform time-series prediction of the vibration signal characteristics and obtain the predicted value.

[0042] The training of the trend prediction model in this embodiment must meet the following conditions: the training data comes from more than 10 consecutive hours of normal operation data for the corresponding speed zone, to ensure that the model can learn the typical characteristics of the vibration signal of the engine during normal operation in that speed zone. The network structure adopts a two-layer LSTM with 64±20% hidden layer units, and uses INT8 quantization for deployment, which reduces the computational complexity and storage space requirements of the model while ensuring model performance; the Huber loss function is used for training, with the δ parameter set to 1.5±0.3. The Huber loss function has better robustness in handling outliers, improving the training effect and prediction accuracy of the model.

[0043] In addition, during the steady-state operating condition early warning process, if the prediction error of the trend prediction model of a certain region exceeds twice the historical average prediction error of that region for 10 consecutive times, the retraining process of the trend prediction model of that region will be automatically triggered. The retraining data includes the latest 2 hours of normal operation data and 8 hours of data randomly selected from the original training data to ensure the prediction accuracy of the model.

[0044] Step S400: Residual analysis.

[0045] In this step, the absolute value of the difference between the characteristics of the collected aero-engine vibration signal and the predicted value of the trend prediction model is calculated to obtain the absolute residual e. t The absolute residual e t This reflects the degree of difference between measured and predicted data, through the analysis of the absolute residual e. t The analysis can determine whether there are any abnormalities in the engine's operating status.

[0046] Step S500: Dynamic baseline management.

[0047] This step updates the mean μ of the absolute residuals of the current partition based on the sliding window statistical method. t and standard deviation σ t This establishes a dynamic baseline. This step employs a dual-window mechanism for baseline management, specifically as follows: Figure 3 As shown.

[0048] For a short window (60±15s), designed to capture and identify sudden failures, this implementation uses the Exponentially Weighted Moving Average (EWMA) algorithm to calculate the mean μ of the absolute residuals of the current partition. t and standard deviation σ t Mean μ t The calculation formula is:

[0049] μ t =αe t +(1-α)μ t-1 ;

[0050] The weighting coefficient α = 0.2 is used to enhance the responsiveness to recent data changes.

[0051] For the long-term window (1 ± 0.2 h), which aims to identify slowly developing incremental faults, this implementation uses a fixed window length and records the historical maximum residual e. max When a long-term window is updated, new data is added to the window while the oldest data is removed. When the current absolute residual e... t >0.8*e max When the baseline is triggered, an emergency update is performed. The updated baseline will be applied to subsequent residual comparisons to ensure that the baseline can adapt to long-term changes that may occur during engine operation, thus guaranteeing the accuracy and effectiveness of the baseline.

[0052] Step S600: Tiered early warning.

[0053] In this step, a three-level early warning mechanism is implemented based on the degree of deviation between the residuals and the baseline, as follows:

[0054] Level 1 warning: When e t >μ t +3σ t When an e-signal occurs, a Level 1 warning is triggered. t >μ t +3σ t As an abnormal event, abnormal events are continuously accumulated, and no other additional processing is required when a Level 1 warning is issued;

[0055] Level 2 warning: When abnormal events accumulate to 10 consecutive times, a Level 2 warning is triggered. When a Level 2 warning is triggered, a local alarm needs to be issued, and maintenance operators are notified to pay attention to the operating status of the aircraft engine and potential faults through flashing lights, sound prompts, and other means.

[0056] Level 3 Warning: When e t >μ t +6σ t When the fault is detected, a Level 3 warning is triggered directly. When a Level 3 warning is detected, protective actions are activated to prevent the fault from worsening.

[0057] It is easy to see that using a pre-trained trend prediction model to predict the characteristic time series of vibration signals can deeply learn the time series characteristics of vibration signals, and has higher prediction accuracy than traditional methods. Combined with dynamic baseline management, the baseline is updated in real time to adapt to changes in engine operating status. Through a three-level early warning mechanism, the fault level is accurately judged based on the degree of deviation between the residual and the baseline, which significantly improves the sensitivity and reliability of fault detection and effectively reduces the false alarm rate and the missed alarm rate.

[0058] The second embodiment of the present invention relates to an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the first embodiment of the air-engine vibration fault early warning method based on operating condition adaptation.

[0059] The third embodiment of the present invention relates to a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the first embodiment of the air-engine vibration fault early warning method based on operating condition adaptation.

[0060] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0061] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0062] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction methods implemented in a process. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0063] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0064] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An aero-engine vibration fault early warning method based on working condition self-adaption, characterized in that, The method comprises the following steps: Synchronously collecting vibration signals and rotating speed data of an aero-engine during operation through vibration sensors and rotating speed sensors; Calculating an instantaneous rotating speed change rate based on the collected rotating speed data, and determining an aero-engine working condition according to the instantaneous rotating speed change rate; When the aero-engine working condition is a steady-state working condition, discretizing the rotating speed data into N partitions, and using a trend prediction model corresponding to the partition to predict a vibration time sequence feature, thereby obtaining a predicted value; Calculating an absolute residual of a feature of the collected vibration signals of the aero-engine and the predicted value, and performing baseline management and hierarchical early warning based on the absolute residual.

2. The working condition adaptive aero-engine vibration fault early warning method according to claim 1, characterized in that, The determination of the aero-engine working condition according to the instantaneous rotating speed change rate is specifically: when an absolute value of a difference between a current rotating speed and a previous rotating speed divided by a sampling interval is greater than or equal to a preset value, the aero-engine is in a transient working condition; when the absolute value of the difference between the current rotating speed and the previous rotating speed divided by the sampling interval is less than the preset value, the aero-engine is in a steady-state working condition.

3. The working condition adaptive aero-engine vibration fault early warning method according to claim 1, characterized in that, When the rotating speed data is discretized into N partitions, non-uniform discretization processing is adopted, and 1000 RPM is taken as an interval for partitioning in a low rotating speed interval, and 500 RPM is taken as an interval for partitioning in a high rotating speed interval, wherein the low rotating speed interval and the high rotating speed interval are continuous intervals.

4. The working condition adaptive aero-engine vibration fault early warning method according to claim 1, characterized in that, The trend prediction model corresponding to the partition takes the rotating speed data of the partition as input, and takes the feature of the vibration signals of the aero-engine as output. The training data of the trend prediction model corresponding to the partition is derived from normal operation data of the corresponding partition for more than K hours.

5. The working condition adaptive aero-engine vibration fault early warning method according to claim 4, characterized in that, Further, when a prediction error of the trend prediction model corresponding to the partition exceeds twice a historical average prediction error of the partition for M times continuously, a retraining process of the trend prediction model corresponding to the partition is triggered, and the retraining data comprises p hours of newly collected normal operation data and K-p hours of randomly extracted data from original training data.

6. The method of claim 1, wherein, The baseline management is specifically: updating a mean value and a standard deviation of the absolute residual of the current partition based on a sliding window statistical method, and establishing a dynamic baseline based on the mean value and the standard deviation; For a short-term window, an exponential weighted moving average algorithm is used to calculate the mean value and the standard deviation of the absolute residual of the current partition; for a long-term window, a fixed window length is adopted, a historical maximum absolute residual is recorded, and when the absolute residual of new data exceeds a preset multiple of the historical maximum absolute residual, baseline updating is triggered.

7. The aero-engine vibration fault early warning method based on working condition self-adaption according to claim 6, wherein The hierarchical early warning is specifically: implementing a three-level early warning mechanism according to a deviation degree of the absolute residual from the baseline, and specifically as follows: when the absolute residual is greater than a sum of the mean value and 3 times the standard deviation, triggering a first-level early warning, determining an abnormal event, and continuously accumulating the abnormal event; when the abnormal event is continuously accumulated to a preset number of times, triggering a second-level early warning, and prompting a maintenance operator to pay attention to the operating state of the aero-engine through a prompt; when the absolute residual is greater than a sum of the mean value and 6 times the standard deviation, triggering a third-level early warning, and enabling a protection action.

8. The method of claim 1, wherein, The frozen fault early warning mechanism is used when the aero-engine is in a transient state.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The computer program is executed by the processor to implement the steps of the aero-engine vibration fault early warning method based on working condition self-adaption as claimed in any one of claims 1-8.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the aero-engine vibration fault early warning method based on working condition self-adaption as claimed in any one of claims 1-8.

Images