An aero-engine real-time early warning method, system, device, medium and product
By employing a collaborative architecture of airborne edge computing and ground-based deep temporal reconstruction analysis, combined with an adaptive threshold determination mechanism, the real-time performance and detection sensitivity issues of aero-engine performance monitoring have been resolved, enabling acute identification and efficient early warning of early faults.
Patent Information
- Application Number
- CN202611100774.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-23
- Publication Date
- 2026-11-10
AI Technical Summary
Existing methods for monitoring the performance of aero-engines suffer from poor real-time performance, low sensitivity to early signs of failure, high false alarm and false negative rates, and a lack of adaptive capabilities.
An airborne edge computing module is used to segment and extract features from engine parameter data. Combined with a ground-to-air data transmission module, the data is sent to the ground early warning center in real time. The data is then reconstructed using a preset deep time series model. Based on the statistical distribution of the engine's recent historical reconstruction errors, the early warning threshold is dynamically calculated to generate early warning information.
This achieves an improvement from the traditional delay of several hours after the event to a response time of minutes during flight, increasing the sensitivity of detecting early minor performance degradation, reducing false alarm rates, and enhancing the real-time performance and intelligence of engine monitoring.
Smart Images

Figure CN123072152A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of aero-engine technology, and in particular to a real-time early warning method, system, device, medium and product for aero-engines. Background Technology
[0002] Currently, aero-engine performance monitoring primarily relies on post-flight offline analysis of Quick Access Recorder (QAR) data. This involves downloading data after a flight and then using fixed thresholds or simple trend models to identify anomalies. However, this approach suffers from warning lags of several hours to days, failing to respond promptly to nascent faults or early performance degradation. Furthermore, fixed threshold methods struggle to capture the complex nonlinear coupling relationships between multiple parameters and cannot effectively track the slow performance degradation trend of the engine. This leads to the missed detection of early, subtle fault symptoms, while parameter fluctuations within normal ranges are easily misreported as anomalies. In addition, threshold setting and trend judgment are highly dependent on human experience, lacking adaptive and intelligent capabilities, making it difficult to meet the demands of real-time, accurate, and intelligent monitoring. Summary of the Invention
[0003] This invention provides a real-time early warning method, system, device, medium, and product for aero-engines to solve the technical problems of poor real-time performance, low sensitivity to early fault signs, high false alarm and missed alarm rates, and lack of adaptive capability in traditional aero-engine performance monitoring methods.
[0004] According to one aspect of the present invention, a real-time early warning method for an aero-engine is provided, applied to an aero-engine real-time early warning system, the system including an airborne edge computing module, an air-to-ground data transmission module, and a ground-based early warning center; the method includes: The airborne edge computing module continuously receives engine parameter data, segments and extracts features from the engine parameter data to obtain the time-series feature data corresponding to each time segment. The time-series characteristic data are sequentially sent to the ground early warning center via the air-to-ground data transmission module. The ground-based early warning center reconstructs each time-series feature data sequentially based on a preset deep time-series model, and generates an early warning information when multiple consecutive reconstruction errors exceed the early warning threshold. The early warning threshold is dynamically calculated based on the statistical distribution of the engine's recent historical reconstruction errors.
[0005] According to another aspect of the present invention, a real-time early warning system for an aircraft engine is provided, the system comprising: The airborne edge computing module, deployed on the aircraft, is used to continuously receive engine parameter data, segment the engine parameter data and extract features to obtain the time-series feature data corresponding to each time-series segment. The air-to-ground data transmission module connects the airborne edge computing module and the ground early warning center, and is used to send various time-series characteristic data to the ground early warning center in sequence; The ground-based early warning center is used to reconstruct each time-series feature data sequentially based on a preset deep time-series model, and to generate early warning information when multiple consecutive reconstruction errors exceed the early warning threshold. The early warning threshold is dynamically calculated based on the statistical distribution of the engine's recent historical reconstruction errors.
[0006] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the real-time early warning method for aircraft engines according to any embodiment of the present invention.
[0007] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the real-time early warning method for aero-engines according to any embodiment of the present invention.
[0008] According to another aspect of the present invention, a computer program product is provided, the computer program product comprising a computer program that, when executed by a processor, implements the real-time early warning method for aero-engines according to any embodiment of the present invention.
[0009] The technical solution of this invention continuously receives engine parameter data through an airborne edge computing module, segments and extracts features from the engine parameter data to obtain time-series feature data corresponding to each time segment. The time-series feature data is then sequentially sent to a ground-based early warning center via a ground-to-air data transmission module. The ground-based early warning center reconstructs each time-series feature data sequentially based on a preset deep time-series model, and generates an early warning message when multiple consecutive reconstruction errors exceed an early warning threshold. The early warning threshold is dynamically calculated based on the statistical distribution of the engine's recent historical reconstruction errors. This technical solution, through a collaborative architecture of real-time processing at the airborne edge and deep time-series reconstruction analysis at the ground, combined with a continuous exceedance judgment mechanism and an adaptive threshold based on the engine's recent historical error distribution, achieves an improvement from a traditional delay of several hours after the event to a minute-level response during flight. It also improves the detection sensitivity of early minor performance degradation and effectively suppresses false alarms caused by occasional disturbances.
[0010] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description
[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0012] Figure 1 This is a flowchart of a real-time early warning method for an aircraft engine according to Embodiment 1 of the present invention; Figure 2 This is a flowchart of a real-time early warning method for an aircraft engine according to Embodiment 2 of the present invention; Figure 3 This is a flowchart of another real-time early warning method for aero-engines provided according to Embodiment 2 of the present invention; Figure 4 This is a schematic diagram of the structure of a real-time early warning system for an aircraft engine according to Embodiment 3 of the present invention; Figure 5 This is a schematic diagram of the structure of an electronic device that implements the real-time early warning method for aero-engines according to embodiments of the present invention. Detailed Implementation
[0013] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0014] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0015] Example 1 Figure 1 This is a flowchart of a real-time early warning method for an aircraft engine provided in Embodiment 1 of the present invention. This embodiment is applicable to situations where the performance status of an aircraft engine is monitored online and continuously during flight, and real-time early warnings are provided for early performance degradation and potential faults. This method can be executed by an aircraft engine real-time early warning system, which can be implemented in hardware and / or software and can be configured in electronic devices. Figure 1 As shown in the figure, the real-time early warning method for aero-engines provided in this embodiment is applied to a real-time early warning system for aero-engines. The method specifically includes the following steps: S110. Continuously receive engine parameter data through the airborne edge computing module, and segment and extract features from the engine parameter data to obtain the time-series feature data corresponding to each time-series segment.
[0016] The airborne edge computing module can refer to an embedded computer device or dedicated processing board deployed on the aircraft's airborne end, possessing data acquisition, signal processing, and lightweight artificial intelligence inference capabilities, and capable of connecting to the QAR system bus. Engine parameter data can refer to various monitoring index data used to quantify the current thermodynamic performance, rotor dynamics characteristics, and gas path operating status of an aero-engine, such as, but not limited to, low-pressure rotor speed (N1), high-pressure rotor speed (N2), exhaust gas temperature (EGT), and fuel flow (FF). A time-series segment can refer to a multi-dimensional parameter data block segmented from a continuous engine parameter data stream, containing multiple consecutive sampling time points arranged chronologically, which completely preserves the dynamic evolution trajectory of each parameter of the engine over time within that time period. Time-series feature data can refer to low-dimensional standardized feature vector data of a single time-series segment after lightweight network feature filtering and dimensionality compression, retaining the core performance change patterns of the engine and eliminating redundant noise information.
[0017] In this embodiment of the invention, the airborne edge computing module deployed on the aircraft, acting as a listening node of the airborne data bus, can continuously receive the aircraft parameter data stream broadcast on the bus and read multi-channel engine parameter data, such as N1, N2, EGT, FF, etc. Next, the airborne edge computing module can preprocess the received engine parameter data stream (e.g., time alignment, filtering, standardization, etc.) and segment it to generate multiple independent time-series segments with uniform duration, continuous timing, and no loss of data information. Finally, the airborne edge computing module can call a pre-configured lightweight feature extraction network to perform spatial dimension feature mining on each time-series segment, extracting the linkage change features and operating condition coupling relationships of multiple engine parameters within the segment, filtering out invalid and redundant information, retaining the core effective features that reflect engine performance degradation and operating condition changes, and finally outputting low-dimensional time-series feature data uniquely corresponding to each time-series segment for subsequent air-to-ground transmission and ground-based in-depth analysis.
[0018] S120 transmits each time-series characteristic data sequentially to the ground early warning center via the air-to-ground data transmission module.
[0019] The air-to-ground data transmission module can refer to a dedicated functional module deployed on the airborne end, adapted to the aircraft's narrowband air-to-ground communication link, used to realize bidirectional reliable data transmission between the airborne edge computing module and the ground early warning center. The ground early warning center can refer to an intelligent analysis platform deployed on the ground operation and maintenance service end, equipped with deep time-series models, dynamic threshold judgment, and early warning decision logic, used to complete the detection of aero-engine anomalies and the generation of early warnings.
[0020] In this embodiment of the invention, the air-to-ground data transmission module can receive the time-series feature data output by the airborne edge computing module, and encapsulate the time-series feature data corresponding to each time-series segment into data frames according to the time sequence, and send them to the ground early warning center through the Aircraft Communications Addressing and Reporting System (ACARS), satellite communication link and other means.
[0021] S130. The ground-based early warning center reconstructs each time-series feature data sequentially based on a preset depth time-series model, and generates early warning information when multiple consecutive reconstruction errors exceed the early warning threshold. The early warning threshold is dynamically calculated based on the statistical distribution of the engine's recent historical reconstruction errors.
[0022] The pre-set deep time-series model refers to a deep learning model trained in a supervised or unsupervised manner using a large amount of historical normal engine parameter data. Internally, it can employ an encoder-decoder architecture specifically designed to learn the dynamic evolution patterns of the engine under healthy conditions. Reconstruction error refers to the deviation between the actual input time-series feature data and the reconstruction result output by the model, used to quantify the degree to which the current operating condition deviates from the normal mode. The warning threshold is a critical error value used to determine whether the engine's operating condition is abnormal. When the reconstruction error exceeds this value, it indicates that the current operating state deviates from the normal operating range. Each time-series feature data (or time-series segment) corresponds to its own exclusive dynamic warning threshold. This dynamic warning threshold can be calculated based on the statistical distribution characteristics of the reconstruction error value sequence corresponding to the engine's recent historical time-series features before the current detection time, thus overcoming the inherent defects of traditional fixed threshold methods that cannot take into account individual differences between different engines and the inherent defects of baseline drift caused by the natural slow decline of the same engine's performance over time. Warning information refers to the structured prompt data, including abnormal state and fault correlation information, output by the ground warning center when it determines that the engine has performance abnormalities.
[0023] In this embodiment of the invention, the ground-based early warning center can continuously monitor the communication link with the air-to-ground data transmission module. Each time it receives a time-series feature data point, it inputs it into a pre-configured, preset-depth time-series model for processing, reconstructing a standard time-series feature sequence under normal operating conditions, i.e., the reconstruction result. Then, the original input features and the reconstructed features are compared dimension-by-dimensionally to quantify the reconstruction error corresponding to the current time-series feature data. While calculating the current reconstruction error, the ground-based early warning center can also retrieve the engine's recent (e.g., the last 50 flight segments) historical reconstruction errors and determine the early warning threshold corresponding to the current time-series feature data by statistically analyzing the distribution of these historical reconstruction errors. Then, the ground-based early warning center compares the current reconstruction error with the dynamically calculated early warning threshold. If the current reconstruction error does not exceed the early warning threshold, the system records the current reconstruction error and continues processing the next time-series feature data point; if the current reconstruction error exceeds the early warning threshold, the system marks it as a potential anomaly and retrospectively checks the judgment results of multiple preceding time-series segments. Only when the reconstruction errors of multiple consecutive (e.g., three consecutive) time series segments all exceed their respective dynamic warning thresholds will the system finally confirm that the engine status is abnormal, and then generate structured warning information and push it to the human-machine interface for maintenance engineers to monitor and make decisions.
[0024] The technical solution of this invention uses an airborne edge computing module to perform data acquisition and feature extraction upfront, and combines this with a ground-to-air data transmission module to achieve real-time downlink transmission. This reduces the early warning response time from several hours in traditional post-event analysis to minutes during flight. Simultaneously, a preset deep time-series model is used to reconstruct the input, and the degree of deviation is quantified using reconstruction errors. This allows for the sensitive capture of complex nonlinear co-evolutionary relationships among multiple parameters, enabling early, minor performance degradation to be effectively identified well before reaching the hard threshold, significantly reducing the false alarm rate. Furthermore, by combining an adaptive threshold and a continuous over-limit judgment mechanism dynamically calculated based on the statistical distribution of recent historical reconstruction errors, the system effectively filters out occasional disturbances, significantly reduces the false alarm rate, and adaptively tracks the natural performance degradation trajectory of individual components, comprehensively improving the real-time performance, detection sensitivity, and intelligence level of engine monitoring.
[0025] Example 2 Figure 2 This is a flowchart of a real-time early warning method for an aircraft engine provided in Embodiment 2 of the present invention. It is further optimized and extended based on the above embodiments and can be combined with various optional technical solutions in the above embodiments. For example... Figure 2 As shown in the figure, the real-time early warning method for aero-engines provided in this embodiment includes the following steps: S210: Continuously acquire engine parameter data from the airborne data bus through the airborne edge computing module, and preprocess the engine parameter data.
[0026] The airborne data bus can refer to the standardized data transmission bus of the aircraft's airborne avionics system, such as the ARINC 429 or AFDX bus. The airborne digital flight data acquisition unit (DFDAU) broadcasts various flight and engine parameters in a unified manner, providing a homogeneous and standardized data stream transmission channel for the airborne avionics equipment.
[0027] In this embodiment of the invention, after power-on, the airborne edge computing module can be continuously connected to the airborne data bus as a passive listening node, listening in real time to the full range of engine parameter messages periodically broadcast by DFDAU on the bus, and extracting various engine parameter data for engine performance monitoring. Then, preprocessing operations such as timing alignment, filtering, and standardization can be performed on the captured engine parameter data stream to transform the messy, heterogeneous, and timing-biased raw bus data into a standardized engine timing data stream that is timing-ordered, uniformly distributed, and truly effective, providing reliable data input for subsequent timing segmentation and feature extraction.
[0028] S220: The preprocessed engine parameter data is divided into multiple time segments according to a preset length using the airborne edge computing module.
[0029] The preset length can refer to a uniform window size that is pre-set and fixed during the system design phase to cut the data stream in order to adapt to the fixed dimensional requirements of the input layer of the downstream deep neural network model. This size can be expressed as a fixed number of sampling points or a fixed time span (such as a flight segment). This embodiment does not limit this.
[0030] In this embodiment of the invention, after preprocessing the parameter data stream, the airborne edge computing module can uniformly segment the continuous long-time-series data stream according to a pre-configured preset length. The data stream is split using a continuous sliding segmentation method, decomposing the uninterrupted ultra-long-time-series data into multiple independent short-time-series segments with uniform time spans and consistent structural specifications. Furthermore, before formal segmentation, the airborne edge computing module can simultaneously read the aircraft system's air-to-ground status signals (such as landing gear wheel load signals or ground speed signals) and engine thrust setting signals (such as throttle lever angle). Through these external auxiliary markers, the complete flight physical phase (e.g., from takeoff roll to touchdown) can be identified, ensuring that during subsequent segmentation, the same time-series segment does not span two vastly different physical flight phases (e.g., the climb and descent phases are not mixed in one segment), thereby ensuring that each segment has a relatively consistent operating condition background.
[0031] S230: The airborne edge computing module calls a lightweight convolutional neural network to extract features from each time segment in sequence, thereby obtaining the time-series feature data corresponding to each time segment.
[0032] Lightweight Convolutional Neural Networks (CNNs) refer to lightweight CNN models optimized through structural pruning, parameter compression, and convolution simplification for scenarios with limited computing power and high real-time processing requirements on airborne edges. Compared to standard convolutional neural networks, they have the characteristics of small number of parameters, fast operation speed, and low computing power consumption, making them suitable for real-time feature extraction operations on airborne devices.
[0033] In this embodiment of the invention, the airborne edge computing module can input the segmented and integrity-verified standard time-series segments one by one into the locally deployed lightweight CNN. The CNN captures the spatial coupling and correlation features of different engine parameters within the same time-series window, mines the implicit performance rules of the coordinated changes between parameters, discards invalid noise data and redundant parameter information, and finally outputs the time-series feature data corresponding to each time-series segment.
[0034] This embodiment solves the problem of high computational load and inability to perform real-time inference on airborne systems by deploying and calling a lightweight convolutional neural network on the airborne end for pre-feature extraction. At the same time, by multi-parameter spatial feature mining and dimensionality compression, redundant noise information in the original parameters is eliminated, which greatly reduces the amount of data transmitted in subsequent air-to-ground data transmission. Thus, while fully preserving the core features of aero-engine performance changes, it also takes into account the real-time performance of data processing, high transmission efficiency, and accurate feature extraction, providing high-quality standardized feature data support for high-precision time-series reconstruction and early warning of performance anomalies in the cloud.
[0035] S240 transmits each time-series characteristic data sequentially to the ground early warning center via the air-to-ground data transmission module.
[0036] S250. After receiving each time series feature data, the ground early warning center inputs the time series feature data into a preset deep time series model for reconstruction, determines the reconstruction error between the time series feature data and the reconstruction result, and determines the early warning threshold corresponding to the time series feature data based on a preset sliding window mechanism; wherein, the preset deep time series model adopts an encoder-decoder structure.
[0037] Among them, the preset sliding window mechanism can refer to the dynamic data statistics mechanism adopted by the ground early warning center. It uses the current time series feature data to be detected as the time base, and extracts a fixed number of historical time series data samples to form a sliding statistical window. The window is updated segment by segment over time and is used to statistically analyze the error distribution characteristics in real time.
[0038] In this embodiment of the invention, the ground-based early warning center can employ a real-time streaming monitoring mechanism to receive time-series feature data sequentially transmitted from the air-to-ground data transmission module. Each received time-series feature data item can be input into a preset deep time-series model for reconstruction. This preset deep time-series model can employ a Long Short-Term Memory (LSTM) network or a Transformer model with an encoder-decoder structure. Specifically, the encoder is used to mine the long-term temporal dependencies and normal operating conditions of the time-series feature data, compressing and generating latent feature vectors. The decoder is used to reverse-decode and restore the latent feature vectors, outputting a reconstructed feature sequence that is dimensionally consistent and temporally aligned with the input time-series feature data—the reconstruction result. Next, the original input time-series feature data and the reconstruction result generated by the model are compared dimensionally and time-point by time, quantifying the overall deviation between the two sets of data to obtain the reconstruction error corresponding to the current time-series segment, thus accurately reflecting the degree of deviation between the current engine operating state and historical normal operating conditions. Meanwhile, in order to determine whether the current reconstruction error is abnormal, the system does not use a fixed threshold, but instead uses a preset sliding window mechanism to dynamically determine the warning threshold corresponding to the current time segment.
[0039] In one embodiment, determining the early warning threshold corresponding to the time-series feature data based on a preset sliding window mechanism specifically includes the following steps: Based on the chronological order, a predetermined number of historical time series feature data points preceding the time series feature data are selected to form the historical error window corresponding to the time series feature data. Based on the statistical distribution of each historical reconstruction error in the historical error window, the warning threshold corresponding to the time series feature data is determined.
[0040] Among them, statistical distribution can refer to the numerical distribution pattern of all historical reconstruction errors within the historical error window. For example, it can include statistical characteristics such as error mean, standard deviation, and dispersion, which can objectively reflect the error fluctuation range under the recent normal operating conditions of the engine.
[0041] In this embodiment of the invention, the process of dynamically determining the warning threshold specifically includes: S1. The ground-based early warning center uses the currently processed time-series feature data as the latest time node. Following chronological order, it selects multiple sets of historical time-series feature data that are consecutive and adjacent in time before the generation of the current time-series feature data. Then, for each selected set of historical time-series feature data, it retrieves the corresponding historical reconstruction error values that have already been calculated, summarizing them to obtain a complete and continuous set of historical error sample data. The summarized historical reconstruction errors of a preset number are then integrated to form the historical error window corresponding to the current time-series feature data. It can be understood that as new time-series feature data is continuously streamed in, this historical error window will automatically slide forward, removing the earliest historical error samples and incorporating the latest historical error samples, thus achieving dynamic window updates.
[0042] S2 can perform statistical analysis on all historical reconstruction errors within the historical error window, solve for the mean, standard deviation and other statistical characteristics of the error data within the window, thereby accurately depicting the overall fluctuation range and distribution pattern of the reconstruction error under the recent normal operating conditions of the engine, and eliminating the individual adaptation deviation caused by static global statistics.
[0043] S3. Based on the statistical distribution characteristics of historical error windows, a specific early warning threshold adapted to the current engine's recent operating status can be calculated, ensuring that each time-series feature data has an independent judgment standard that fits the real-time operating conditions, replacing the traditional fixed and uniform threshold. For example, the formula for calculating the dynamic early warning threshold can be: Adaptive early warning threshold = Average window error + Preset confidence coefficient K × Standard deviation of window error, where the preset confidence coefficient K is an adjustable hyperparameter greater than zero, used to balance the detection sensitivity and false alarm tolerance of the early warning system.
[0044] This embodiment introduces a statistical distribution dynamic threshold mechanism based on a sliding historical window, which completely eliminates the inherent defects of the traditional aircraft-wide fixed threshold and global static statistical threshold used in aero-engine early warning. It can accurately adapt to the individual manufacturing differences of a single engine, the real-time fluctuations of different flight conditions, and the slow performance degradation throughout the entire life cycle. This ensures that the anomaly judgment criteria for each time segment are consistent with the engine's recent actual operating status, thereby effectively reducing false alarms and missed alarms caused by operating condition drift and equipment aging, and significantly improving the accuracy and adaptability of early performance degradation and weak anomaly identification of aero-engines.
[0045] Furthermore, the pre-set deep time-series model can be trained in advance using only historical normal engine parameter data. By allowing the model to learn only the standard feature patterns of healthy engine operation, the interference of fault samples and abnormal data on model feature learning is completely avoided. This effectively solves the industry pain points of traditional supervised learning models, such as reliance on fault samples, poor adaptability to small-sample faults, and confusion between normal and abnormal operating condition features. It significantly improves the model's sensitivity to early subtle performance degradation and latent anomalies of the engine, accurately distinguishes between normal operating condition fluctuations and real performance anomalies, and significantly reduces the false alarm and missed alarm rates, thereby ensuring the accuracy and reliability of real-time anomaly warnings for aero-engines.
[0046] For example, consider monitoring high-pressure compressor efficiency degradation. A pre-set deep time-series model learns the coordinated variation patterns of parameters such as N2 speed, EGT, and FF under normal conditions during the cruise stabilization phase. When high-pressure compressor efficiency begins to decline slightly, EGT may show a slight upward trend and FF may increase slightly in order to maintain the same thrust. This slight temporal variation, deviating from the normal coordinated pattern, is captured by the deep time-series model, manifesting as a continuous small increase in reconstruction error, thus triggering a low-level warning in the early stages of performance degradation, rather than waiting for the absolute value of EGT to exceed a hard threshold.
[0047] S260. When the ground-based early warning center detects that the reconstruction errors of multiple consecutive time-series feature data all exceed their respective early warning thresholds, it determines that the time-series segment corresponding to the current time-series feature data is an abnormal segment.
[0048] In this embodiment of the invention, for the currently processed time-series feature data, the ground-based early warning center can compare the reconstruction error corresponding to the time-series segment with its dedicated dynamic early warning threshold. If the current reconstruction error does not exceed the warning threshold, the system records the current reconstruction error and continues processing the next time-series feature data. If the current reconstruction error exceeds the warning threshold, the system marks it as a potential anomaly and retrospectively checks the judgment results of multiple preceding time-series segments. After confirming that the reconstruction errors of multiple consecutive time-series segments all exceed the limit, the ground-based early warning center determines the time-series segment corresponding to the latest received current time-series feature data as the engine operation anomaly segment, completing the accurate location of the abnormal operating condition and providing effective data basis for subsequent fault analysis and early warning generation.
[0049] S270. The ground-based early warning center generates corresponding early warning information based on abnormal segments.
[0050] In this embodiment of the invention, after locating the abnormal segment, the ground-based early warning center can retrieve the flight condition information corresponding to the abnormal segment. By comprehensively analyzing the fault information, it can generate early warning information containing key information such as the fault level, the flight segment where the abnormality occurred, the suspected faulty component, and maintenance recommendations. Furthermore, this early warning information can be pushed to the human-machine interface for monitoring and decision-making by aircraft maintenance engineers.
[0051] In one embodiment, the ground-based early warning center generates corresponding early warning information based on abnormal segments, specifically including the following steps: The analysis results of the flight phase and parameter contribution corresponding to the abnormal segments were determined by the ground-based early warning center. The ground-based early warning center generates early warning information based on the flight phase and parameter contribution analysis results; the early warning information includes at least the anomaly level, suspected faulty components, and maintenance recommendations.
[0052] The flight phase refers to the typical operational intervals of an aircraft from takeoff to landing, defined by thrust status and configuration changes, such as takeoff, climb, cruise, descent, and approach. Parameter contribution analysis results refer to the percentage of error contribution of each original engine parameter, derived by the ground-based early warning center from anomaly time-series characteristic data, used to quantify and locate the core parameters causing overall operational anomalies. Anomaly level refers to the engine performance anomaly hierarchy classified based on factors such as the magnitude and duration of the anomaly, used to distinguish between different degrees of anomaly, such as minor performance degradation, moderate anomaly, and severe failure risk. Suspected faulty components refer to engine anomaly-related parts identified by matching core anomaly parameters with the component operating characteristics of the corresponding flight phase, achieving precise location from parameter anomalies to component failures. Maintenance recommendations refer to targeted inspection, troubleshooting, and maintenance guidance plans generated for suspected faulty components and anomaly types, providing aircraft maintenance personnel with directly implementable operational guidelines.
[0053] In this embodiment of the invention, the process of generating early warning information specifically includes: S1. After determining that a certain time segment is an abnormal segment, the ground early warning center can immediately retrieve the flight phase tag carried by the segment during airborne preprocessing, compare the tag with the preset flight phase division standard library, and accurately confirm the specific phase in which the anomaly occurred, such as the takeoff run, initial climb, cruise stabilization, step descent, or landing deceleration phase.
[0054] S2. The early warning center initiates a parameter attribution analysis program to identify the root causes of the reconstruction error. This is done by performing a reverse chain reaction, layer by layer, along the network structure of the pre-defined deep time-series model, to obtain the parameter contribution analysis results. The core logic of this calculation process is to calculate the degree of influence (i.e., sensitivity gradient) on the final reconstruction error value when each independent sensor parameter (such as EGT, N2, FF, etc.) in the input data undergoes a small change. Through this backtracking calculation, the system assigns a quantitative contribution score to each input parameter, which directly reflects the responsibility weight that parameter bears for the abnormal deviation from the normal pattern.
[0055] S3. The ground early warning center combines the magnitude of the abnormal segment reconstruction error exceeding the dynamic adaptive early warning threshold, the number of consecutively exceeding time series segments, and the duration of the abnormality, and classifies the engine performance abnormality according to the preset classification judgment rules to distinguish different levels such as slight performance degradation, moderate operational abnormality, and serious fault risk, so as to achieve quantitative classification of the degree of abnormality.
[0056] S4. Based on the matched flight phase operating condition characteristics and the selected core abnormal parameter change patterns, the ground early warning center retrieves a pre-set aero-engine operating condition-parameter-fault association knowledge base. It then matches the component degradation and fault characteristics corresponding to parameter anomalies under different operating conditions, accurately identifying the suspected faulty component corresponding to this anomaly. For example, a slow increase in EGT and a slight increase in FF during the cruise phase typically indicates high-pressure turbine blade performance degradation.
[0057] S5: The ground-based early warning center can match the determined anomaly level and the suspected faulty component with the corresponding standardized operation and maintenance rules and inspection standards to generate targeted fault investigation, component testing, and preventive maintenance suggestions. Finally, it integrates key information such as anomaly level, suspected faulty component, and dedicated maintenance suggestions to generate complete structured early warning information and complete the output and push.
[0058] This embodiment addresses the shortcomings of traditional aero-engine early warning systems, which can only output single anomaly alerts, lack fault location, and provide no maintenance guidance. Through a linkage judgment mechanism that adapts to flight phase operating conditions and quantifies parameter contribution, it achieves differentiated fault determination based on operating conditions, accurate tracing of core anomaly parameters, precise location of faulty components, and quantitative classification of anomaly levels. This effectively avoids the problems of misjudgment and missed judgment of faults under different flight conditions, significantly improving the accuracy and precision of engine anomaly early warning. It can provide aircraft maintenance personnel with intuitive, accurate, and actionable maintenance decision-making basis, reduce blind troubleshooting and ineffective maintenance, effectively ensure the safe operation of aero-engines, and improve equipment maintenance efficiency.
[0059] The technical solution of this invention utilizes an airborne edge computing module to complete data acquisition, segmentation, and feature extraction using a lightweight convolutional neural network during flight, significantly compressing data transmission volume. Combined with real-time downlink transmission via a ground-to-air data link, this reduces the early warning response time from hours / days in traditional post-event analysis to minutes during flight, achieving a shift from reactive post-event monitoring to proactive real-time early warning. The ground-based early warning center employs a deep temporal model with an encoder-decoder structure for reconstruction. By exploiting reconstruction errors, it keenly captures the complex nonlinear co-evolutionary patterns among multiple parameters, enabling effective identification of early, minor performance degradation before reaching physical hard thresholds, significantly improving detection sensitivity. Simultaneously, a sliding window mechanism dynamically calculates corresponding adaptive thresholds, and a judgment mechanism that triggers an alarm only when multiple consecutive errors exceed the threshold effectively filters out occasional spikes caused by airflow disturbances or sensor noise, significantly reducing false alarm rates while maintaining a high detection rate.
[0060] Figure 3 This is a flowchart of another real-time early warning method for aero-engines provided in Embodiment 2 of the present invention. Figure 3 As shown, this scheme acquires QAR sensor data in real time through an airborne edge computing module and performs time synchronization, filtering, denoising, and standardization. The data is then segmented into time-series segments by flight segment, and spatial features are extracted and compressed using a lightweight CNN. The compressed feature data is then transmitted to the ground-based early warning center in real time via ACARS or satellite link. The ground-based early warning center uses a pre-trained encoder-decoder deep temporal model (LSTM / Transformer) trained only with normal data to reconstruct each segment and calculate the reconstruction error. Then, for each segment, an adaptive dynamic threshold is calculated based on the engine's recent historical error statistical distribution. An anomaly is only identified when the errors of multiple consecutive segments exceed their respective thresholds. Finally, structured early warning information, including anomaly level, suspected faulty components, and maintenance recommendations, is generated and pushed out in real time, combining flight phase and parameter contribution. This process shortens the early warning response time from the traditional hours after the event to minutes during flight. It uses the sensitivity of deep models to multi-parameter nonlinear collaborative modes to capture early and subtle signs of degradation. At the same time, it effectively filters out occasional disturbances and significantly reduces false alarms through continuous over-limit and dynamic threshold mechanisms. It also provides traceable root cause analysis support for maintenance decisions, significantly improving the real-time performance, accuracy and intelligence of engine condition monitoring.
[0061] Example 3 Figure 4 This is a schematic diagram of the structure of a real-time early warning system for an aircraft engine provided in Embodiment 3 of the present invention. Figure 4 As shown, the system includes: The airborne edge computing module 31 is deployed on the aircraft to continuously receive engine parameter data, segment the engine parameter data and extract features to obtain the time-series feature data corresponding to each time-series segment. The air-to-ground data transmission module 32 connects the airborne edge computing module 31 and the ground early warning center 33, and is used to send various time-series characteristic data to the ground early warning center 33 in sequence. Ground-based early warning center 33 is used to reconstruct each time-series feature data sequentially based on a preset depth time-series model, and generate early warning information when multiple consecutive reconstruction errors exceed the early warning threshold; wherein, the early warning threshold is dynamically calculated based on the statistical distribution of the engine's recent historical reconstruction errors.
[0062] Furthermore, based on the above embodiments of the invention, the airborne edge computing module 31 is specifically used for: Engine parameter data is continuously acquired from the airborne data bus and preprocessed. The preprocessed engine parameter data is divided into multiple time segments according to a preset length; A lightweight convolutional neural network is invoked to extract features from each time segment in sequence, thereby obtaining the time feature data corresponding to each time segment.
[0063] Furthermore, based on the above embodiments of the invention, the ground early warning center 33 is specifically used for: After each received temporal feature data, the temporal feature data is input into a preset deep temporal model for reconstruction. The reconstruction error between the temporal feature data and the reconstruction result is determined, and the warning threshold corresponding to the temporal feature data is determined based on a preset sliding window mechanism. The preset deep temporal model adopts an encoder-decoder structure. When the reconstruction errors of multiple consecutive time series feature data are detected, all of which exceed their respective warning thresholds, the time series segment corresponding to the current time series feature data is determined to be an abnormal segment. Generate corresponding warning information based on abnormal segments.
[0064] Furthermore, based on the above embodiments of the invention, when the ground early warning center 33 determines the early warning threshold corresponding to the time-series feature data based on a preset sliding window mechanism, it is specifically used for: Based on the chronological order, a predetermined number of historical time series feature data points preceding the time series feature data are selected to form the historical error window corresponding to the time series feature data. Based on the statistical distribution of each historical reconstruction error in the historical error window, the warning threshold corresponding to the time series feature data is determined.
[0065] Furthermore, based on the above embodiments of the invention, when the ground early warning center 33 generates corresponding early warning information based on abnormal segments, it is specifically used for: Determine the flight phase and parameter contribution analysis results corresponding to the abnormal segments; Based on the analysis results of flight phases and parameter contribution, early warning information is generated; the early warning information includes at least the anomaly level, suspected faulty components, and maintenance recommendations.
[0066] Furthermore, based on the above embodiments of the invention, the preset deep time series model is trained in advance using only historical normal engine parameter data.
[0067] The real-time early warning system for aero-engines provided in this embodiment of the invention can execute the real-time early warning method for aero-engines provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.
[0068] Example 4 Figure 5 A schematic diagram of an electronic device 40 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.
[0069] like Figure 5 As shown, the electronic device 40 includes at least one processor 41 and a memory, such as a read-only memory (ROM) 42 or a random access memory (RAM) 43, communicatively connected to the at least one processor 41. The memory stores computer programs executable by the at least one processor. The processor 41 can perform various appropriate actions and processes based on the computer program stored in the ROM 42 or loaded from storage unit 48 into the RAM 43. The RAM 43 may also store various programs and data required for the operation of the electronic device 40. The processor 41, ROM 42, and RAM 43 are interconnected via a bus 44. An input / output (I / O) interface 45 is also connected to the bus 44.
[0070] Multiple components in electronic device 40 are connected to I / O interface 45, including: input unit 46, such as keyboard, mouse, etc.; output unit 47, such as various types of monitors, speakers, etc.; storage unit 48, such as disk, optical disk, etc.; and communication unit 49, such as network card, modem, wireless transceiver, etc. Communication unit 49 allows electronic device 40 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0071] Processor 41 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 41 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 41 performs the various methods and processes described above, such as the real-time early warning method for aircraft engines.
[0072] In some embodiments, the real-time warning method for aircraft engines can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 48. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 40 via ROM 42 and / or communication unit 49. When the computer program is loaded into RAM 43 and executed by processor 41, one or more steps of the real-time warning method for aircraft engines described above can be performed. Alternatively, in other embodiments, processor 41 can be configured to perform the real-time warning method for aircraft engines by any other suitable means (e.g., by means of firmware).
[0073] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.
[0074] In some embodiments, the real-time early warning method for aircraft engines can be implemented as a computer program, which is implicitly included in a computer program product. When executed by a processor, the computer program implements the real-time early warning method for aircraft engines of the present invention. The computer program product can be understood as a software product that primarily implements its solution through a computer program. The computer program used to implement the method of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer program causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The computer program can be executed entirely on a machine, partially on a machine, partially on a remote machine as a standalone software package, or entirely on a remote machine or server.
[0075] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.
[0076] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).
[0077] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.
[0078] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.
[0079] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.
[0080] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.
Claims
1. A real-time early warning method for aircraft engines, characterized in that, An application is made in a real-time early warning system for aero-engines, the system comprising an airborne edge computing module, an air-to-ground data transmission module, and a ground-based early warning center; the method includes: The airborne edge computing module continuously receives engine parameter data, segments and extracts features from the engine parameter data to obtain time-series feature data corresponding to each time segment. The time-series characteristic data are sequentially sent to the ground early warning center via the air-to-ground data transmission module. The ground-based early warning center reconstructs each of the time-series feature data sequentially based on a preset deep time-series model, and generates early warning information when multiple consecutive reconstruction errors exceed the early warning threshold; wherein, the early warning threshold is dynamically calculated based on the statistical distribution of the engine's recent historical reconstruction errors.
2. The method according to claim 1, characterized in that, The process involves continuously receiving engine parameter data through the airborne edge computing module, segmenting the engine parameter data, and extracting features to obtain time-series feature data corresponding to each time segment, including: The airborne edge computing module continuously acquires engine parameter data from the airborne data bus and preprocesses the engine parameter data. The airborne edge computing module divides the preprocessed engine parameter data into multiple time segments according to a preset length. The airborne edge computing module calls a lightweight convolutional neural network to extract features from each of the time segments in sequence, thereby obtaining the time-series feature data corresponding to each time segment.
3. The method according to claim 1, characterized in that, The process involves the ground-based early warning center sequentially reconstructing each of the time-series feature data based on a preset depth time-series model, and generating early warning information when multiple consecutive reconstruction errors exceed the early warning threshold. This includes: After receiving the time-series feature data, the ground-based early warning center inputs the time-series feature data into the preset deep time-series model for reconstruction, determines the reconstruction error between the time-series feature data and the reconstruction result, and determines the early warning threshold corresponding to the time-series feature data based on a preset sliding window mechanism; wherein, the preset deep time-series model adopts an encoder-decoder structure; When the ground-based early warning center detects that the reconstruction errors of multiple consecutive time-series feature data exceed their respective early warning thresholds, it determines that the time-series segment corresponding to the current time-series feature data is an abnormal segment. The ground-based early warning center generates corresponding early warning information based on the abnormal segments.
4. The method according to claim 3, characterized in that, The step of determining the early warning threshold corresponding to the time-series feature data based on a preset sliding window mechanism includes: According to the chronological order, a preset number of historical time-series feature data points located before the time-series feature data are selected to form the historical error window corresponding to the time-series feature data. Based on the statistical distribution of each historical reconstruction error in the historical error window, the warning threshold corresponding to the time series feature data is determined.
5. The method according to claim 3, characterized in that, The step of generating the corresponding early warning information based on the abnormal fragments through the ground-based early warning center includes: The ground-based early warning center determines the flight phase and parameter contribution analysis results corresponding to the abnormal segment. The ground-based early warning center generates the early warning information based on the flight phase and the contribution of the parameters; the early warning information includes at least the anomaly level, suspected faulty components, and maintenance recommendations.
6. The method according to claim 1, characterized in that, The preset deep time series model is trained in advance using only historical normal engine parameter data.
7. A real-time early warning system for an aircraft engine, characterized in that, The system includes: An airborne edge computing module, deployed on an aircraft, is used to continuously receive engine parameter data, segment the engine parameter data and extract features to obtain time-series feature data corresponding to each time-series segment. The air-to-ground data transmission module connects the airborne edge computing module and the ground early warning center, and is used to send the aforementioned time-series feature data to the ground early warning center in sequence. The ground-based early warning center is used to reconstruct each of the time-series feature data sequentially based on a preset deep time-series model, and generate early warning information when multiple consecutive reconstruction errors exceed the early warning threshold; wherein, the early warning threshold is dynamically calculated based on the statistical distribution of the engine's recent historical reconstruction errors.
8. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the real-time early warning method for an aircraft engine as described in any one of claims 1-6.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are used to cause a processor to execute the real-time early warning method for an aero-engine as described in any one of claims 1-6.
10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the real-time early warning method for aero-engines as described in any one of claims 1-6.
