An aircraft skin door anomaly diagnosis and prediction method

Abstract

The application discloses an aircraft skin valve anomaly diagnosis and prediction method, and relates to the technical field of aircraft operation and maintenance.The application adopts a big data analysis method, is committed to capturing the signs before the valve failure, and establishes an advanced failure prediction model, through which potential failure signs of the skin valve can be recognized in advance, thereby providing strong support for preventive maintenance, so that the possibility of failure can be effectively reduced, the smoothness of fleet operation can be ensured, and more accurate operation and maintenance management means are provided for the airline, and the overall operation efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of aircraft maintenance technology, specifically to a method for diagnosing and predicting abnormalities in aircraft skin valves. Background Technology

[0002] According to A320 fleet operational reliability data, a malfunction in the skin valve of the electronic cabin ventilation system has had a significant impact on flight operations. In a milder case, it may lead to the execution of the M item reserved in the short stop MEL, resulting in flight delays; in a more severe case, it may affect the pressurization system after takeoff, causing the cabin altitude to be lost, thus forcing the aircraft to return to its origin or make an emergency landing.

[0003] Since 2017, data from China Southern Airlines' fleet of unplanned replacements shows that the VFT300B00 skin exit valve, with less than 6,000 hours of service, accounts for a significant proportion of unplanned replacements. This data reveals that even after high-level overhauls, more than half of the unplanned replacement parts still have less than 6,000 hours of service.

[0004] Further analysis revealed that when the skin valve is in one of the three states of not fully open, not fully closed, or partially open, if the valve opening and closing time exceeds 20 seconds, the valve position information will be displayed as XX;

[0005] Therefore, an abnormally long valve opening and closing time is one of the main causes of the malfunction.

[0006] The existing monitoring methods for skinned valves are as follows:

[0007] (1) Time monitoring threshold:

[0008] Set thresholds for the closing and opening times of the skinned valve. Once the valve's opening and closing time reaches the set threshold, the system will directly send a warning and recommend replacing the spare part as soon as possible.

[0009] (2) Trend monitoring:

[0010] The historical data of the valve is smoothed, and the average value of every 10 points is used to represent the smoothed value of the current point.

[0011] Analyze the most recent third of all smoothed values. If the difference between the maximum and minimum values ​​exceeds 0.7, it indicates that the valve travel time is trending upward.

[0012] (3) Monitoring of critical value distribution:

[0013] Consider the state where the valve opening and closing time is consistently long but does not exceed the threshold.

[0014] Analyze the most recent 1 / 3 of the smoothed value data. If the average maximum smoothed value for the opening time reaches 10.5 or the average maximum smoothed value for the closing time reaches 10, it is judged as an abnormal situation.

[0015] Based on the above three types of logic, a comprehensive review of 54 skin-mounted outlet valve malfunctions from 2018 to the present revealed that 29 of these malfunctions could be monitored using these three types of logic, achieving a monitoring efficiency of approximately 53.7%. Therefore, monitoring valve status using these logics can prevent more than half of the malfunctions, which aligns with the team's objectives.

[0016] However, the existing technical solutions mainly have the following problems:

[0017] (1) Current technology has a relatively low hit rate in detecting skin valve anomalies, which is insufficient to meet the needs of daily operations. We urgently need a more accurate and efficient method to ensure the timely detection and accurate assessment of skin valve anomalies. This is crucial for improving the efficiency of aircraft maintenance and flight safety;

[0018] (2) During the testing process, too many relevant parameters need to be compared and analyzed, resulting in relatively low efficiency. This not only increases the time cost of testing, but may also reduce the accuracy of testing in some cases. We urgently need a more intelligent and efficient testing method that can accurately capture skin valve anomalies without excessive parameter comparison, thereby improving testing efficiency, reducing resource waste, and ensuring the reliability of test results;

[0019] Therefore, a new solution is needed to address the above problems. Summary of the Invention

[0020] The purpose of this invention is to provide a method for diagnosing and predicting abnormalities in aircraft skin valves, so as to solve the technical problems mentioned in the background art.

[0021] To achieve the above objectives, the present invention provides the following technical solution: a method for diagnosing abnormalities in aircraft skin valves, comprising at least the following steps:

[0022] S1: Merge multiple HDF5 files from the same flight to generate a corresponding CSV file containing 5 relevant parameters;

[0023] S2: Process the CSV file data to calculate the time required for the valves to open and close during a single takeoff and landing of a flight;

[0024] S3: Use the sliding window method, set the window size to 10, select a smoothing distance of 10 (group the data, and calculate the mean of the data in each window);

[0025] S4: Set the data thresholds. Set the open time threshold to 11.5 seconds and the close time threshold to 11.3 seconds.

[0026] S5: Check for abnormal situations and determine whether the mean of the grouped data exceeds the threshold. If it does, it is considered an abnormal situation.

[0027] A method for predicting aircraft skin valves includes at least the following steps:

[0028] S1: Data smoothing and preprocessing, which uses a sliding window technique to smooth the raw data;

[0029] S2: Model selection and parameter estimation, determining the parameters of the ARIMA model, and selecting parameters based on a preliminary analysis of data characteristics and model fit;

[0030] S3: Model Fitting and Diagnosis. To ensure the adaptability and reliability of the model, the ARIMA model is fitted using the training set, and model diagnosis is performed by comparing the residuals (differences) between the observed values ​​and the model predictions.

[0031] S4: Prediction and estimation. After fitting the ARIMA model, the fitted model is used to predict data at future time points in order to make predictions about the future trend of the data.

[0032] Preferably, S1 includes at least the following steps: setting the size of the sliding window to 10 and the sliding batch to 5, thereby reducing noise and unexpected fluctuations in the data.

[0033] Preferably, determining the parameters of the ARIMA model in step S2 includes at least the following steps:

[0034] By observing the graphs of the autocorrelation function (ACF) and partial autocorrelation function (PACF), the order (p, d, q) of the ARIMA model was initially selected.

[0035] Meanwhile, a first-order difference operation (d=1) was performed on the non-stationary data;

[0036] The ARIMA model with parameters p=3, d=1, and q=0 was then determined.

[0037] Compared with the prior art, the beneficial effects of the present invention are:

[0038] This invention employs big data analytics to capture signs of impending valve failure and establishes an advanced failure prediction model. This model enables us to identify potential malfunctions in skin valves in advance, providing strong support for preventative maintenance and effectively reducing the likelihood of failures. This approach not only helps ensure the smooth operation of the fleet but also provides airlines with more precise operational management tools, improving overall operational efficiency. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the 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.

[0040] Figure 1 This is a schematic diagram showing the overall comparison results of the present invention. Detailed Implementation

[0041] 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.

[0042] Example 1:

[0043] A method for diagnosing abnormalities in aircraft skin valves includes at least the following steps:

[0044] S1: Merge multiple HDF5 files from the same flight to generate a corresponding CSV file containing 5 relevant parameters;

[0045] S2: Process the CSV file data to calculate the time required for the valves to open and close during a single takeoff and landing of a flight;

[0046] S3: Use the sliding window method, set the window size to 10, select a smoothing distance of 10 (i.e., 20 data points, divided into 2 groups) for grouping, and calculate the mean of the data in each window;

[0047] S4: Set the data thresholds. Set the open time threshold to 11.5 seconds and the close time threshold to 11.3 seconds.

[0048] S5: Check for abnormal situations and determine whether the mean of the grouped data exceeds the threshold. If it does, it is considered an abnormal situation.

[0049] The above-described embodiment 1 outlines the data processing and fault identification steps for aircraft HDF5 files, determining the presence of anomalies through specific conditions and thresholds. Values ​​exceeding the threshold average are identified as anomalies, indicating abnormal valve opening or closing times.

[0050] Comparison of differences

[0051] When comparing the current monitoring methods with our proposed method, it is noteworthy that the current methods require three different monitoring techniques to detect anomalies in the skinned valve. In contrast, our proposed method only requires one method—sliding window threshold monitoring—to effectively detect anomalies.

[0052] A significant difference lies in the set monitoring thresholds. In our sliding window threshold monitoring, we set fairly strict standards for the opening and closing times of the valves: an opening time threshold of 11.5 seconds and a closing time threshold of 11.3 seconds. In contrast, the currently used critical value distribution monitoring method sets different time standards, for example, an opening time of 10.5 seconds and a closing time of 10 seconds.

[0053] Our proposed method significantly improves the results, especially in early warning, increasing the warning rate from 69% to 100%. This greatly enhances the accuracy and efficiency of early warning. The comparison results are as follows: Figure 1 As shown;

[0054] It's worth noting that the sliding window threshold monitoring method, by analyzing the sliding window and mean of the data, can observe the timing of the valve opening and closing in greater detail. Compared to other methods, this approach provides a more comprehensive and accurate means of anomaly detection, helping to capture subtle changes in the valve's state and identify potential problems more quickly.

[0055] Experimental conclusion:

[0056] An analysis of 16 aircraft and over 20,000 flights revealed 11 abnormal aircraft, achieving a 100% early warning rate and an 80% effective warning rate. Therefore, our method can improve the early warning rate for flights.

[0057] Our method has demonstrated significant effectiveness, particularly in early warning, with the warning rate soaring from 69% to 100%. This improvement substantially enhances the accuracy and usability of our early warning system. Notably, the sliding window threshold monitoring method offers more precise anomaly detection capabilities. It can observe changes in the skin valve's state in greater detail and quickly identify potential problems. This method not only effectively captures subtle changes in valve status but also detects and resolves potential faults in advance, thus providing a more reliable guarantee for flight safety.

[0058] Example 2:

[0059] A method for predicting aircraft skin valves includes at least the following steps:

[0060] S1: Data smoothing and preprocessing, which uses a sliding window technique to smooth the raw data;

[0061] S2: Model selection and parameter estimation, determining the parameters of the ARIMA model, and selecting parameters based on a preliminary analysis of data characteristics and model fit;

[0062] S3: Model Fitting and Diagnosis. To ensure the model's adaptability and reliability, the ARIMA model was fitted using the training set. Model diagnosis was performed by comparing the residuals (differences) between the observed values ​​and the model's predicted values. Furthermore, we conducted an autocorrelation test on the model's residual sequence to ensure that the residual sequence was random and without significant autocorrelation, a crucial characteristic of an effective model.

[0063] S4: Prediction and estimation. After fitting the ARIMA model, the fitted model is used to predict data at future time points in order to make predictions about the future trend of the data.

[0064] To evaluate the model's predictive performance, we will use a test dataset to validate its accuracy. We can use the root mean square error (RMSE) or other metrics to assess the difference between the model's predictions and the actual observations, thereby evaluating the model's predictive power and accuracy. This evaluation process is an important test of the model's effectiveness in real-world applications. It is worth noting that excellent models typically have an RMSE less than 0.2. An RMSE below this value indicates a good fit to the data, with a small difference between the predicted results and the actual observations, which is generally considered a sign of good model performance.

[0065] Based on the model's predictive capabilities, we can use it to predict two future time steps. Considering we used a sliding window with a sliding batch size of 5, the model can predict the status changes of up to five flights in advance. This means we can use the model to detect abnormal flight statuses in advance. By training the model and predicting data at future time points, we have the ability to capture potential flight anomalies within a certain period, which helps to take timely measures to address possible problems or anomalies. This predictive capability provides valuable early warning capabilities for flight status monitoring and management.

[0066] S1 includes at least the following steps: setting the sliding window size to 10 and the sliding batch size to 5 (i.e., moving 5 data points each time), thereby reducing noise and unexpected fluctuations in the data. The sliding window technique helps to establish a smooth trend in time series, making the data more stable and thus better reflecting the overall pattern and trend.

[0067] Determining the parameters of the ARIMA model in S2 includes at least the following steps:

[0068] By observing the graphs of the autocorrelation function (ACF) and partial autocorrelation function (PACF), the order (p, d, q) of the ARIMA model was initially selected.

[0069] Meanwhile, a first-order difference operation (d=1) was performed on the non-stationary data;

[0070] The ARIMA model with parameters p=3, d=1, and q=0 was then determined.

[0071] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for diagnosing abnormalities in aircraft skin valves, characterized in that: At least the following steps are included: S1: Merge multiple HDF5 files from the same flight to generate a corresponding CSV file containing 5 relevant parameters; S2: Process the CSV file data to calculate the time required for the valves to open and close during a single takeoff and landing of a flight; S3: Use the sliding window method, set the window size to 10, select a smoothing distance of 10 for grouping, and calculate the mean of the data in each window; S4: Set the data thresholds. Set the open time threshold to 11.5 seconds and the close time threshold to 11.3 seconds. S5: Check for abnormal situations and determine whether the mean of the grouped data exceeds the threshold. If it does, it is considered an abnormal situation.

2. A method for predicting aircraft skin valves, characterized in that: At least the following steps are included: S1: Data smoothing and preprocessing, which uses a sliding window technique to smooth the raw data; S2: Model selection and parameter estimation, determining the parameters of the ARIMA model, and selecting parameters based on a preliminary analysis of data characteristics and model fit; S3: Model Fitting and Diagnosis. To ensure the adaptability and reliability of the model, the ARIMA model is fitted using the training set, and model diagnosis is performed by comparing the residuals between the observed values ​​and the model predictions. S4: Prediction and estimation. After fitting the ARIMA model, the fitted model is used to predict data at future time points in order to make predictions about the future trend of the data.

3. The prediction method for aircraft skin valves according to claim 2, characterized in that: The S1 includes at least the following steps: setting the size of the sliding window to 10 and the sliding batch to 5, thereby reducing noise and unexpected fluctuations in the data.

4. The method for diagnosing and predicting abnormalities in aircraft skin valves according to claim 2, characterized in that: Determining the parameters of the ARIMA model in S2 includes at least the following steps: By observing the graphs of the autocorrelation function (ACF) and partial autocorrelation function (PACF), the order (p, d, q) of the ARIMA model was initially selected. Meanwhile, a first-order difference operation (d=1) was performed on the non-stationary data; The ARIMA model with parameters p=3, d=1, and q=0 was then determined.

Images