ACM data analysis method and data monitoring system

Abstract

The invention discloses an ACM data analysis method and a data monitoring system, and belongs to the technical field of aircraft maintenance, and the method comprises the steps: constructing an ACM data set, setting a feature engineering processing method, carrying out coupling feature extraction through setting a dynamic window and calculating a dynamic correlation coefficient, dividing working conditions, eliminating fluctuation interference, and generating a structured time series data set; the method comprises the steps of obtaining ACM reference data of aircrafts of the same model, setting a reference fault judgment method, judging latest data in a structured time sequence data set, screening abnormal parameters, judging whether the ACM breaks down or not by combining an actual temperature adjusting effect, reducing single-parameter misjudgment, setting a coupling fault positioning method, diagnosing the faulty ACM, and distinguishing true faults and false anomalies. Meanwhile, a prediction scheduling method is set, fault prediction, health state prediction and service life evaluation are carried out on the fault-free ACM, the probability of sudden faults is reduced, and non-planned flight stopping is reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of aircraft maintenance, and relates to an ACM data analysis method and a data monitoring system. Background Art

[0002] The air cycle machine (ACM) in the aircraft air conditioning system plays a key role in ensuring a suitable environment in the cabin and is the core component of aircraft environmental control. The ACM is mainly composed of a compressor and a turbine cooler. The turbine cooler undertakes a crucial heat exchange task. It exchanges heat with the outside air or other media and cooperates with the compressor to cool and regulate the cabin air.

[0003] The existing Chinese patent with authorization announcement number CN107807628B discloses a method for evaluating the performance degradation of a heat exchanger in a civil aircraft air conditioning system, including: first collecting key performance parameters, then establishing a heat exchanger monitoring parameter system around the key performance parameters, and then using the monitoring parameter system under a fault-free state to establish a heat exchanger performance baseline model. When the system receives new monitoring parameters, the heat exchanger performance baseline model calculates an estimated value of the key performance parameter, and then subtracts the estimated value from the actual monitoring value to obtain a deviation value of the monitoring parameter. The deviation value is monitored and analyzed, and when the characteristic value is abnormal, a fault warning is issued.

[0004] Although the prior art solves the problem of difficulty in online monitoring of heat exchangers in air conditioning systems, avoids unplanned maintenance of air conditioning systems caused by heat exchanger failures, reduces unnecessary cleaning and replacement of heat exchangers, saves maintenance costs for airlines, and improves aircraft utilization, it lacks comprehensive consideration of the coupling of multiple components of aircraft air conditioning systems (especially ACM) and health management throughout the life cycle, resulting in insufficient accuracy of fault diagnosis and lack of predictability in maintenance strategies. In the ACM of aircraft air conditioning systems, the compressor and turbine cooler work in conjunction. If the dynamic interaction between the compressor and turbine cooler is not considered, it will lead to misjudgment of faults, resulting in inaccurate fault location and incorrect maintenance, such as misdiagnosing compressor problems as heat exchanger failures. At the same time, since aviation maintenance requires advance planning of spare parts and grounding cycles, if only fault warnings are performed without considering the use of historical deviation data to predict the remaining service life, the proactive maintenance needs cannot be met, making it impossible for airlines to optimize spare parts inventory and flight scheduling. Therefore, the present application provides an ACM data analysis method and a data monitoring system. Summary of the invention

[0005] Aiming at the deficiencies of the existing technology, the purpose of the present invention is to provide an ACM data analysis method and a data monitoring system. By means of a dynamic window, the coupling characteristics of the compressor and the turbine cooler are extracted, and the Granger causality analysis is used to accurately locate the coupling faults and avoid misjudgment. At the same time, combined with a deep autoencoder and a survival analysis model, the remaining life is predicted based on historical data, providing support for proactive maintenance, optimizing spare parts inventory and flight scheduling, and realizing the coupling effect of multiple components and the health management of the entire life cycle.

[0006] To achieve the above object, the present invention provides the following technical solutions: The ACM data analysis method includes: Construct an ACM data set, set the feature engineering processing method, extract the coupling characteristics by calculating the dynamic correlation coefficient through setting a dynamic window, and divide the working conditions to generate a structured time series data set; Obtain the ACM benchmark data of the same type of aircraft, set the benchmark fault judgment method, judge the latest data in the structured time series data set, screen abnormal parameters, and combine the actual temperature adjustment effect to judge whether the ACM is faulty. Set the coupling fault location method to diagnose the faulty ACM, distinguish true faults from pseudo-abnormalities, and at the same time set the prediction scheduling method to predict faults for the non-faulty ACM, and conduct health status prediction and life assessment.

[0007] Furthermore, when generating the structured time series data set, the feature engineering processing method includes: Read the time stamps of the ACM data set, calculate the difference between adjacent time stamps, and define the mode of the difference as the time resolution; Conduct a response test on the ACM, simulate the operation under different working conditions, monitor the fluctuation characteristics of key parameters, and determine the time range when the key parameters exceed the preset change range; Dynamically adjust the window size and sliding step based on a preset proportional coefficient; Slide the window along the time series, conduct Pearson correlation analysis on the compressor parameters and turbine cooler parameters in each window to generate a dynamic correlation coefficient; Associate the dynamic correlation coefficient with the start time stamp of the corresponding window to form a coupling feature time series.

[0008] Furthermore, the feature engineering processing method also includes: Divide the climb, cruise, and descent stages according to the flight parameters, traverse the ACM data set, and mark the flight stage of the data at each time point based on the preset flight stage judgment rules; Based on the recognition results of the flight stage and the dynamic correlation coefficient, save the data in the ACM data set and the coupling feature time series to three subsets of climb, cruise, and descent constructed respectively. Standardize the data of each subset; Using the timestamp as the row index, arrange the key parameters and dynamic correlation coefficients into a two-dimensional matrix to generate a structured time series dataset.

[0009] Furthermore, when judging whether the ACM is faulty, the benchmark fault judgment method includes: Obtain the operation data of the same type of aircraft without faults in the initial stage of leaving the factory, screen the data in the cruise stage based on the flight stage judgment rule, and generate a benchmark database; Calculate the mean and standard deviation of the key parameters in the benchmark database, and construct a parameter benchmark interval using the 3σ principle; Integrate the mean of each key parameter and the parameter benchmark interval to construct a standard ACM performance model; Read the latest data in the structured time series dataset, input it into the standard ACM performance model, compare it with the parameter benchmark interval, identify abnormal parameters, and construct an abnormal parameter set and a normal parameter set.

[0010] Furthermore, the benchmark fault judgment method also includes: Read the normal parameters in the normal parameter set, calculate the derivatives between adjacent data points in the continuous sliding window in turn, and obtain the derivative sequence within the window; For any single sliding window, calculate the mean and variance of the derivative sequence for dynamic trend detection; If the mean of the derivative sequence is not within the preset normal mean range or the variance exceeds the preset normal variance range, mark that there is an abnormal derivative change of the normal parameter within the sliding window; otherwise, there is no abnormal derivative change of the normal parameter; Perform abnormal judgment on the continuous sliding window; if there are abnormal derivative changes in all continuous sliding windows, update the normal parameter to an abnormal parameter and transfer it to the abnormal parameter set; otherwise, the normal parameter is not abnormal; Calculate the actual temperature adjustment effect in the cabin and calculate the temperature difference , and comprehensively judge whether the ACM has a fault; When or , it is determined that the ACM has a fault; when and , it is determined that the ACM has no fault; where is the number of abnormal parameters, is the temperature difference threshold, is the abnormal threshold.

[0011] Furthermore, when diagnosing a faulty ACM, the coupled fault location method includes: In the structured time series dataset, extract the data segment corresponding to the latest timestamp according to the timestamp information; Use the Granger causality analysis method to construct a bivariate regression model based on the compressor parameters and the turbine cooler parameters, and evaluate the lag prediction ability between variables; Use the AIC criterion to calculate the AIC values for different lag orders and select the one that minimizes the AIC value as the optimal lag order; Estimate the parameters of the bivariate regression model based on the least squares method, calculate and minimize the sum of squared residuals to fit the data, and calculate the causal relationship strength between the compressor and turbine cooler parameters; Judge whether there is a causal association between the parameters through Granger causality test; if , then determine that the turbine cooler parameter is the Granger cause of the compressor parameter; if , then determine that the compressor parameter is the Granger cause of the turbine cooler parameter; where , are the parameters of the bivariate regression model.

[0012] Furthermore, the coupling fault location method further includes: Based on the result of the Granger causality test, judge whether there is a causal association between the compressor and turbine cooler parameters; If there is no causal relationship, determine that the fault type is a pseudo anomaly and perform pseudo anomaly processing; If there is a causal relationship, judge whether the dynamic correlation coefficient exceeds the preset normal correlation range; If , determine that the fault type is a pseudo anomaly and perform pseudo anomaly processing; where is the Z-score value of the correlation coefficient within the sliding window, is the normal correlation threshold; If , combine the flight phase information to judge the fault type; if the aircraft is in the cruise phase, determine that the fault type is a true fault and perform true fault processing, otherwise, determine that the fault type is a pseudo anomaly and perform pseudo anomaly processing; Generate a maintenance report according to the fault type.

[0013] Furthermore, the true fault processing includes: Based on the abnormal parameter set, obtain the components corresponding to the abnormal parameters in the abnormal parameter set to determine the candidate fault components; Call the historical maintenance case library to screen the historical cases related to the candidate fault components; Calculate the cosine similarity between the current fault data and each historical case data; Sort the historical cases according to the cosine similarity, and output the top 3 fault causes and confidence levels with the highest similarity; Generate a fault location report including the faulty component, cause, and confidence level, where the confidence level is the cosine similarity value.

[0014] Furthermore, the pseudo - anomaly processing includes: Extract sensor data from the structured time - series dataset, check the data validity and continuity point by point, and calculate the continuity index; Compare the continuity index with a preset continuity threshold. If the continuity index is lower than the continuity threshold, mark the corresponding sensor as a suspected fault; otherwise, the sensor is normal; Analyze the noise level of the sensor data, compare the calculated noise index with a preset noise threshold. If the noise index exceeds the noise threshold, mark the sensor as to be calibrated; otherwise, the sensor is normal; Construct a pseudo - anomaly pattern library, store the characteristic data of historical pseudo - anomaly events and manage them by classification; Use the dynamic time warping algorithm to match the sensor data without anomalies with the patterns in the pseudo - anomaly pattern library, and calculate the matching distance; If the matching distance is less than a preset distance threshold, output the corresponding pseudo - anomaly type; Otherwise, update the pseudo - anomaly to a real fault and perform real - fault processing;

[0015] Generate a fault location report including the sensor status and pseudo - anomaly type.

[0016] Furthermore, the ACM data analysis method also includes using a fault prediction method to determine whether the ACM fails; the fault prediction method includes: Construct a deep auto - encoder model, train and evaluate the deep auto - encoder model using the structured time - series dataset, and calculate the reconstruction error based on the output of the deep auto - encoder model; Based on historical fault data, set a degradation weight for each component, and sum the reconstruction errors of each component weighted by their degradation weights to construct a health index; Construct an index prediction model, train the index prediction model with historical health index data, and predict the trend of the health index for future time steps, where the length of the input sequence is and the prediction step size is ; Input the latest health index data points into the index prediction model, and output the future Predicted values of health indicators for a time step; Construct a survival analysis model based on historical failure data, input the predicted values of the health indicators and key parameters into the survival analysis model, and calculate the predicted failure probabilities at different time points; Compare the predicted failure probabilities at different time points with a preset failure threshold. Once it is detected that the predicted failure probability at a certain time point exceeds the failure threshold, it is determined that the ACM is about to fail, and the remaining life of the ACM is output.

[0017] A data monitoring system, including: a data reading module, a diagnostic analysis module, a prediction module, and a feedback module; The data reading module is used to construct an ACM data set, set a dynamic window, extract coupled features based on the dynamic correlation coefficient, and divide the working conditions to generate a structured time series data set; The diagnostic analysis module is used to screen out abnormal parameters by combining static thresholds and parameter change rates, and combine the actual temperature adjustment effect to judge whether the ACM is faulty, and diagnose the faulty ACM to distinguish true faults from pseudo anomalies; The prediction module is used to predict potential faults of the fault-free ACM, and judge whether the ACM fails according to the fault prediction result, and calculate the remaining life of the ACM; The feedback module is used to receive the feedback result after the staff performs maintenance operations to judge the accuracy of fault diagnosis and prediction, and perform update operations.

[0018] Advantages of the present invention: Through dynamic window and flight phase division, the dynamic coupling relationship between parameters is effectively captured, the interference of working condition fluctuations on data analysis is eliminated, and the accuracy of feature extraction is improved; the benchmark fault judgment combines static thresholds and dynamic trend detection, which can quickly identify obvious anomalies and capture short-term parameter fluctuations, improving the sensitivity and reliability of fault detection. And the coupling fault location uses Granger causality analysis and decision trees to quantify the interactive effects of components, solves the misjudgment problem of traditional single-parameter threshold methods, and realizes the accurate location of coupling faults; the health indicator prediction model based on deep autoencoders and LSTM can detect potential fault trends in advance; combined with the survival analysis model, the remaining life can be accurately estimated, providing a basis for preventive maintenance; the maintenance feedback mechanism continuously optimizes model parameters and algorithms by comparing diagnosis and prediction results, forming a closed-loop optimization system, continuously improving the accuracy and adaptability of the system, while ensuring flight safety, significantly reducing maintenance costs and improving aircraft operation efficiency. Description of the Drawings

[0019] Figure 1 It is a flowchart of the ACM data analysis method; Figure 2It is a flowchart of the feature engineering processing method of the present invention; Figure 3 It is a flowchart of the reference fault judgment method of the present invention; Figure 4 It is a flowchart of the coupled fault location method of the present invention; Figure 5 It is a flowchart of the fault prediction method of the present invention; Figure 6 It is a structure diagram of the data monitoring system. Specific implementation mode

[0020] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. Without conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0021] Embodiment 1

[0022] Reference Figures 1 to 5 As shown, this embodiment introduces the ACM data analysis method, including: Connect a dedicated data reading device to the flight parameter card of the aircraft. The flight parameter card, also known as the flight data recording card, is a portable storage medium in the avionics system used to store the flight parameter data of the aircraft. According to the pre-defined data protocol, extract the offline data related to the ACM and flight parameter data, such as the rotational speed of the compressor, the inlet and outlet pressures, the temperature and flow rate of the turbine cooler. At the same time, with the help of database query statements, such as SQL queries, screen out the maintenance record data of the ACM from the aircraft maintenance database, including historical fault types, maintenance times, and replaced parts. Obtain the flight environmental data, such as the outside air temperature, air pressure, and humidity, by establishing a data interface with the airport meteorological station, and use the GPS time or flight phase marker, such as takeoff / landing time; use the GPS time or flight phase marker (such as takeoff / landing time) as a reference to match the timestamps of the environmental data and the offline data. For data with inconsistent timestamps, use the linear interpolation method for processing to align the two in time, generating an ACM dataset with aligned timestamps, ensuring data time alignment and standardization, providing highly reliable input for subsequent analysis. Preprocess the data in the ACM dataset, set the feature engineering processing method, set a dynamic window, calculate the dynamic correlation coefficient of the compressor parameters and the turbine cooler parameters for coupled feature extraction, and at the same time, cluster the data according to the flight phase to divide the working conditions, eliminate the fluctuation interference, and finally generate a structured time series dataset; because during the flight, adjusting the data acquisition method of the on-board equipment or connecting new data acquisition equipment will damage the stability and reliability of the original aircraft system, posing a serious threat to flight safety. At the same time, according to the strict requirements of airworthiness certification, real-time data acquisition belongs to a major modification of the original aircraft design and requires a new comprehensive airworthiness certification, which is time-consuming and costly. In addition, the on-board equipment interfaces and data formats of different models of aircraft are different, making it difficult to develop adapted acquisition equipment and software, and data transmission also faces problems such as real-time performance, stability, and anti-interference. Therefore, considering the above factors, it is more appropriate to adopt the method of processing offline data; Through multi-dimensional data fusion and intelligent algorithms, achieve accurate identification and location of ACM faults; obtain the benchmark data of the ACM of the same model aircraft, set the benchmark fault judgment method, judge the latest data in the structured time series dataset, combine the static threshold and parameter change rate analysis, screen out abnormal parameters, and combine the actual temperature adjustment effect to judge whether the ACM is faulty, reducing single-parameter misjudgment. And set a coupled fault location method to diagnose the faulty ACM, use Granger causality analysis and decision tree algorithms to quantify the interaction effects between components, distinguish true faults from pseudo-abnormalities, improve the accuracy of coupled fault diagnosis, and accurately locate the faulty components, deeply analyze the cause of the fault, classify the fault according to the severity of the fault, and send corresponding maintenance instructions for different levels of faults to notify the maintenance personnel for maintenance; Through health indicator modeling and multi-model fusion, potential faults of a fault-free ACM are predicted, and the remaining life of the faulty component is judged based on the fault prediction results; obtain a fault-free ACM, set a fault prediction method, calculate the predicted health indicators of the ACM to determine whether the ACM fails, calculate the remaining life of the ACM, reduce the probability of sudden faults, extend the service life of the equipment, reduce unplanned flight cancellations, and improve flight operation efficiency; Receive the feedback results after the staff perform maintenance operations to judge the accuracy of fault diagnosis and prediction; for fault diagnosis, compare whether the actual maintenance results are consistent with the diagnosis results. If not, identify the specific differences and make targeted updates and modifications; for fault prediction, compare the prediction with the actual fault time or flight. If the predicted fault time or flight differs significantly from the actual situation, update the model parameters, improve the algorithm, or add new influencing factor variables to improve the accuracy of fault diagnosis and prediction, continuously improve the performance of monitoring the ACM, enhance the adaptability to complex fault modes, ensure the accuracy and robustness of the monitoring system, and form a closed-loop improvement mechanism.

[0023] Furthermore, when generating a structured time series dataset, the feature engineering processing methods include: Different time scales can reflect the operating characteristics of different levels of the ACM. Fixed windows are difficult to comprehensively capture this information. Dynamic windows can flexibly adjust the window size and sliding step according to the actual operating conditions of the ACM, so as to more accurately capture the dynamic relationship of compressor and turbine cooler parameters at different time scales; among them, the size of the window determines the data range considered when calculating the dynamic correlation coefficient, and the sliding step of the window determines the interval at which the window moves in the time series, affecting the frequency of dynamic correlation coefficient calculation and the data coverage; Read the timestamps of the data in the ACM dataset, calculate the difference between adjacent timestamps through the diff function, and obtain the mode of the differences with the mode function to determine the data acquisition frequency, which is defined as the time resolution of the ACM dataset At the same time, conduct a response test on the ACM. By simulating the operation of the ACM under different working conditions, monitor the changes of key parameters, analyze the fluctuation characteristics of key parameters at different time scales, and determine the time range when the key parameters exceed the preset change range Among them, the key parameters include compressor parameters, turbine cooler parameters, and other comprehensive parameters (such as the pressure ratio at the inlet and outlet of the turbine or the pressure ratio at the inlet and outlet of the compressor, the temperature difference at the inlet and outlet of the turbine cooler or the temperature difference with other components), and the values of the key parameters are the offline data in the ACM dataset; Set the proportionality coefficient based on the characteristics and empirical values of the ACM 、 、 、 , and calculate the window size and the sliding step , the expressions are as follows: ; ; Start the window sliding operation from the starting position of the time series data in the ACM dataset. Take the first data point as the starting point of the window, and select data segments according to the window size as the first window data for calculating the dynamic correlation coefficient. Then, slide the window backward along the time series in accordance with the step Each time after sliding, a new window data is obtained, and finally a series of window data for calculating the dynamic correlation coefficient is output; For each window data, based on the Pearson correlation analysis, calculate the dynamic correlation coefficient between the compressor parameters and the turbine cooler parameters within the window. Associate the dynamic correlation coefficient of each window with the corresponding window start timestamp to generate associated data containing the dynamic correlation coefficient and the corresponding window start timestamp; Arrange the dynamic correlation coefficient values in the associated data in chronological order to generate a coupled feature time series, retaining the time order to reflect the dynamic changes of ACM; According to the characteristics and experience of the flight process, formulate flight phase judgment rules. Among them, the flight phases include climb, cruise, and descent. Traverse the flight parameter data in the ACM dataset to identify the flight phase of the aircraft. According to the judgment rules, judge the flight phase of the data at each time point. After determining the flight phase corresponding to a certain time point, find the data record with the same timestamp in the coupled feature time series and add the corresponding flight phase label, and output the coupled feature time series with flight phase labels; Create three independent data subsets, named climb dataset, cruise dataset, and descent dataset respectively. According to the flight phase identification results, classify all the data including the coupled features in the ACM dataset and the coupled feature time series into three phases: climb, cruise, and descent, and store them in the corresponding data subsets; The coupled feature refers to the dynamic correlation coefficient, which reflects the collaborative working relationship between the key components of ACM. Clustering based on the coupled feature helps to gather the data with the same flight phase and similar working conditions together, making the subsequent data analysis more targeted. The data in each subset has similar working condition conditions, effectively reducing the interference of different working condition fluctuations on data analysis; Due to the large differences in the numerical ranges and dimensions of different parameters, for the data within each data subset, the Z-score standardization method is used for processing, and three data subsets after standardization processing are output; Using timestamp as row index, compressor parameters, turbine cooler parameters, dynamic correlation coefficients and other comprehensive parameters are arranged in column order. In each row, the parameter values ​​corresponding to the same time point are filled into the corresponding columns in sequence to construct a two-dimensional matrix. It ensures that each row in the matrix represents a data record at a time point and each column represents a specific parameter, thereby generating a structured time series data set.

[0024] Furthermore, the flight phase judgment rules are as follows: Set the height change threshold to , , the duration time is , ,in, , ; When the altitude change rate is greater than , and lasts longer than , the speed gradually increases (the first-order derivative of the speed is greater than 0), and the engine thrust is in a high power output state (greater than 80% of the cruise thrust), which is determined to be the climbing stage; When the rate of change of altitude is range and lasts longer than , when the speed standard deviation is less than a specific threshold and the engine thrust is stable within the error range of ±10% of the cruise thrust, it is determined to be in the cruise stage; When the altitude change rate is less than , and lasts longer than The speed gradually decreases, and the engine thrust gradually decreases to less than 50% of the cruise thrust until the aircraft lands and stops rolling. This whole process is considered the descent phase, which includes the descent preparation, descent process, and landing roll phase. For special flight situations, such as circling in the air, encountering meteorological disasters or aircraft failures that cause abnormal changes in flight parameters, they are classified according to the trends of the main parameters; if the trends of the main parameters are in line with the characteristic range of the cruise phase, they are classified as the cruise phase; if the parameter change characteristics are closer to the climb or descent phase, they are classified accordingly.

[0025] Furthermore, when determining whether the ACM is faulty, the baseline fault determination method includes: Through the flight parameter card, at least The complete operation data of a trouble-free flight. Since the flight conditions of the aircraft are relatively stable during the cruise phase, the ACM operation status is also relatively stable. The data during this phase can better reflect the parameter characteristics of the normal operation of the system, providing a reliable data basis for building an accurate standard model. Based on the judgment rules of different flight phases, the data of the cruise phase are screened out to generate a benchmark database. For the key parameters in the benchmark database , calculate the mean and the standard deviation , and use the 3σ principle to construct a parameter benchmark interval for the key parameters ; where is the th key parameter in the benchmark database, and , is the number of key parameters in the benchmark database; Integrate the mean and benchmark interval of each key parameter to construct a standard ACM performance model, which is used as the standard for subsequent judgment of whether the latest data is abnormal, to quickly identify parameters that deviate significantly from the normal range. The input of the model is the offline data of each key parameter, and the output is the preliminary judgment result of each key parameter, including abnormal and normal. Compare the input parameter value with the corresponding benchmark interval in the standard ACM performance model. If the parameter value is within the benchmark interval, the output result is normal, and the key parameter is defined as a normal parameter; if the parameter value exceeds the benchmark interval, the output result is abnormal, and the key parameter is defined as an abnormal parameter; Read the latest data in the structured time series dataset and input it into the standard ACM performance model for preliminary anomaly judgment. Construct an abnormal parameter set and a normal parameter set according to the output result, and quickly identify parameters that deviate significantly from the normal range through static threshold detection, providing a basis for preliminary judgment of whether the system is abnormal; among them, the abnormal parameter set is used to save abnormal parameters and the corresponding offline data, and the normal parameter set is used to save normal parameters and the corresponding offline data; Read the normal parameters in the normal parameter set. For each normal parameter, select the latest consecutive sliding windows. In each sliding window, calculate the derivative between adjacent data points in the window in turn to reflect the change rate of the parameter, and obtain the derivative sequence in the window; For each sliding window, analyze the change of the derivative in the window, calculate the mean and variance of the derivative sequence, compare the calculated derivative mean and variance with the pre-set normal fluctuation range data, and analyze the change of the derivative in the sliding window through dynamic trend detection, which helps to capture the abnormal changes of the parameter in the short term, make up for the deficiency of static threshold detection, and improve the sensitivity and accuracy of anomaly detection; If the mean of the derivative sequence is not within the normal mean range, or the variance of the derivative sequence exceeds the normal variance range, mark that there is an abnormal change in the derivative of the corresponding normal parameter in the window; otherwise, there is no abnormality in the normal parameter; Judge the consecutive sliding windows. If the consecutive If there is an abnormal change in the derivative within each sliding window, it is determined that the corresponding normal parameter is abnormal, and the corresponding normal parameter is updated to an abnormal parameter and migrated to the abnormal parameter set; otherwise, the corresponding normal parameter is not abnormal; The ultimate goal of the air conditioning system is to provide a suitable temperature environment for the aircraft cabin. Therefore, the actual temperature adjustment effect is an important indicator to measure whether the ACM is operating normally. By calculating the temperature difference, the refrigeration effect of the system can be verified from the perspective of actual application, avoiding misdiagnosis caused by parameter misjudgment; Obtain the actual temperature adjustment effect in the aircraft And the expected temperature adjustment effect And calculate the temperature difference The expression is as follows: ; In the formula, several temperature values are obtained through temperature sensors installed at various positions inside the aircraft cabin. By calculating the average value of these several temperature values, the actual temperature adjustment effect is obtained, and , is the number of temperature sensors, is the temperature value collected by the Define the key parameters in the abnormal parameter set as abnormal parameters, and obtain the number of abnormal parameters Set the temperature difference threshold to and the abnormal threshold to to determine whether the ACM has a fault, and combine the number of parameter abnormalities and the actual temperature adjustment effect to avoid misjudgment of a single indicator; When or , it is determined that the ACM has a fault; when and , it is determined that the ACM has no fault.

[0026] Furthermore, when diagnosing a faulty ACM, the coupled fault location method includes: When the ACM has a fault, in the structured time series dataset, according to the timestamp information, extract all data segments corresponding to the latest timestamp to obtain the complete operating state data at the moment of the fault, providing comprehensive information for subsequent in-depth analysis of the fault cause, including compressor parameters, turbine cooler parameters, other comprehensive parameters, dynamic correlation coefficients, and flight phase marking information; By quantifying the interactive effects of components, the problem of easy misjudgment in the traditional single-parameter threshold method is solved, and the accurate positioning of coupled faults is achieved. The traditional single-parameter threshold method only focuses on whether a single parameter exceeds the normal range and ignores the interaction between components. Through Granger causality analysis and decision tree construction, considering the relationship between multiple parameters comprehensively helps to more accurately locate coupled faults; using the multivariate time series association rule mining method, such as Granger causality analysis, to construct a bivariate regression model to evaluate the predictive ability of the past values of one variable for the current value of another variable. To determine the optimal lag order of the model, the Akaike information criterion (AIC) is introduced. The AIC criterion can find a balance between model complexity and goodness of fit, ensure the accuracy of the model, and use the AIC criterion to determine the lag order. , for different lag orders calculate the AIC value and select the one that makes the AIC value the smallest as the optimal lag order. The expression is as follows: ; ; In the formula, is the value of the compressor variable at time is the value of the turbine cooler variable at time , , , are the parameters to be estimated, , are independent white noise error terms, is the number of parameters in the bivariate regression model, is the sample size, is the likelihood function, is the index variable representing the lag order, , is at time and also represents the lag period of is at time and also represents the lag period of Use the least squares method to estimate the parameters of the bivariate regression model, calculate the residual sum of squares RSS, and by minimizing the residual sum of squares, make the bivariate regression model accurately fit the data and determine the parameters to be estimated. To quantify the degree of mutual influence between compressor and turbine parameters and calculate the strength of causal relationship , the expression is as follows: ; To determine whether there is a causal relationship between the parameters of the compressor and the turbine cooler, and thus discover potential fault coupling relationships, Granger causality test is carried out; if , it means that the lag value of the turbine cooler variable makes a significant contribution to the prediction of the current value of the compressor variable, then it is determined that the turbine cooler variable is the Granger cause of the compressor variable; if , it means that the lag value of the compressor variable makes a significant contribution to the prediction of the current value of the turbine cooler variable, then it is determined that the compressor variable is the Granger cause of the turbine cooler variable; Based on the verification results of Granger causality, a preliminary judgment on the fault type is made to determine whether there is a Granger causal relationship between the compressor and the turbine cooler. Among them, the causal relationship includes that the turbine cooler variable is the Granger cause of the compressor variable and the compressor variable is the Granger cause of the turbine cooler variable; If there is no causal relationship, indicating that there is no obvious mutual influence between the compressor and turbine cooler parameters, and the fault occurs independently or is caused by other non-coupling factors, then the fault type is determined as a pseudo-anomaly and pseudo-anomaly processing is carried out; If there is a causal relationship, judge whether the dynamic correlation coefficient exceeds the preset normal correlation range; if , it means that the dynamic correlation coefficient exceeds the normal correlation range and is marked as abnormally correlated; if , it means that the dynamic correlation coefficient is within the normal correlation range, the fault type is determined as a pseudo-anomaly and pseudo-anomaly processing is carried out; among them, is the Z-score value of the correlation coefficient within the sliding window, is the normal correlation threshold; When it is determined that the dynamic correlation coefficient is abnormally correlated, the fault type is judged in combination with the flight phase information. Since the aircraft operates stably during the cruise phase and the parameter fluctuations are relatively small, the true fault is determined by analyzing the data during the cruise phase to avoid the influence of transient interference during takeoff and landing on the fault judgment. If the aircraft is in the cruise phase, the fault type is determined as a true fault and true fault processing is carried out. By matching historical maintenance cases, the faulty component and cause are determined; if the aircraft is not in the cruise phase, considering the complex operating environment of the aircraft during takeoff and landing and frequent parameter fluctuations, the anomaly at this time is caused by transient interference, and the fault type is determined as a pseudo-anomaly and pseudo-anomaly processing is carried out; According to the preset fault classification rules, classify the fault handling results and generate a maintenance report, including a fault location report, level, and type; among them, fault handling includes true fault handling and pseudo - anomaly handling, and the output of fault handling is a fault location report.

[0027] Furthermore, true fault handling includes: Based on the abnormal parameter set, obtain the components corresponding to the abnormal parameters and define them as candidate fault components. Call the historical maintenance case library and screen out the historical maintenance cases related to the current candidate fault components, thereby greatly narrowing the data matching range and improving the processing efficiency; among them, the historical maintenance case library stores various fault cases that have occurred in the past ACM, and each case contains detailed data when the fault occurred, such as the parameters of each component, fault phenomena, maintenance measures, and maintenance results. For the screened historical cases, convert the current fault data and historical case data into vector form and calculate the cosine similarity between the current fault data and each case data. Sort the calculated cosine similarities, and output the top 3 fault causes and their corresponding confidence levels. Here, the confidence level is the cosine similarity value. Output a fault location report, including the fault component, cause, and fault confidence level. The fault cause uses historical maintenance experience to quickly locate the fault cause and improve the efficiency and accuracy of fault diagnosis.

[0028] Furthermore, pseudo - anomaly handling includes: Extract sensor data from the structured time - series dataset, check the validity and continuity of the data point by point, and calculate the sensor data continuity index. , to quantify the continuity of the data and determine whether the sensor is working properly; the expression is as follows: ; In the formula, is the total number of sampling points in the structured time - series dataset, is the number of consecutive valid points in the structured time - series dataset. Consecutive valid points are data points detected as normal by the standard ACM performance model; Set the continuous threshold as , and determine whether the sensor is working properly; if , it means that the continuity index exceeds the continuous threshold, there are many discontinuities or abnormal points in the sensor data, the sensor fails, and the sensor is marked as a suspected fault; if , the sensor continuity is normal; Analyze the noise level of the sensor data. Using statistical analysis methods, process the sensor data over a period of time, obtain the standard deviation and mean of the data, and calculate the ratio of the standard deviation to the mean of the data to obtain the noise index. , the expression is as follows: ; In the formula, is the mean value of the sensor data, is the standard deviation of the sensor data; If the noise index exceeds the preset noise threshold, it indicates that the sensor data has excessive noise, which interferes with the fault diagnosis. The sensor is defined as a sensor to be calibrated, and the noise abnormality is recorded; otherwise, the sensor noise is normal; Obtain historical pseudo - abnormal events, extract the key features of different pseudo - abnormal patterns, and convert them into standard pattern data, which are stored in the constructed pseudo - abnormal pattern library. During the storage process, each pseudo - abnormal pattern is classified and managed, a unique identifier is assigned to it, and the characteristic information of the pattern, the applicable sensor type, and the possible causes are recorded in detail to provide a reference for subsequent pattern matching; Using the DTW (Dynamic Time Warping) algorithm, match the sensor data without abnormality with the patterns in the pseudo - abnormal pattern library. By constructing a two - dimensional matrix, calculate the distance between the current data and the pattern data in the library, and find the optimal matching path between the two time series to minimize the cumulative distance on the matching path; if the matching distance is less than the preset distance threshold, it is determined that the current data matches the pseudo - abnormal pattern in the library, and the corresponding pseudo - abnormal type is output; otherwise, the pseudo - abnormal is updated to a real fault, and real - fault processing is performed; among them, the matching distance is the cumulative distance on the optimal matching path; Output a fault location report, including the judgment result of the sensor status, the pseudo - abnormal type (if the matching is successful) or the start information of the real - fault processing process.

[0029] Furthermore, the fault classification rules are as follows: If there is only one abnormal parameter in the abnormal parameter set, it is determined as a minor fault. At this time, the impact on the overall performance of the ACM is small, and manual inspection is arranged before the next flight of the aircraft. Continuously monitor the parameter changes to ensure that the fault will not develop and deteriorate; among them, the next flight of the aircraft is the next complete flight mission planned to be executed after the end of the current flight mission; Multiple parameter abnormalities or parameter abnormalities resulting in a significant decline in the performance of the ACM are determined as moderate faults. At this time, flight safety is not affected, but the parameters of multiple components deviate from the normal range, affecting the cooling or heating effect of the air - conditioning system. A flight suspension for maintenance is arranged within 3 days, and the flight altitude or speed is restricted to reduce the potential risk of the fault to flight safety, and a comprehensive overhaul is carried out; The failure of key functions, such as the complete failure of the compressor or turbine cooler, is determined as a severe fault. At this time, the aircraft air - conditioning system cannot work properly, threatening flight safety. Immediately stop the flight to replace the components and prohibit continued flight to ensure flight safety.

[0030] Further, when determining whether the ACM fails, the fault prediction method includes: When there is no fault in the ACM, a deep autoencoder model is constructed, including an encoder and a decoder. The encoder maps high-dimensional original data, such as compressor parameters and turbine cooler parameters in a structured time series dataset, to a low-dimensional feature space through a multi-layer neural network to extract the key features of the data; the decoder then reconstructs the low-dimensional features into the original data dimension, and the number of layers and the number of neurons in each layer of the model need to be adjusted according to the data characteristics and task requirements; Based on the structured time series dataset, a training set and a test set are divided, and the deep autoencoder model is trained using the training set to minimize the reconstruction error ; In the formula, is the number of samples in the training set, is the original data, is the reconstructed data; ≤ ; Use the trained deep autoencoder model to reconstruct the data in the test set, calculate the reconstruction error, to evaluate the data reconstruction ability of the deep autoencoder model and reflect the degree of data abnormality; There are differences in the importance and degradation speed of different components in the ACM. By setting degradation weights and calculating health indicators, the health state of the ACM can be more accurately evaluated, providing an effective indicator basis for remaining life prediction; based on historical fault data and expert experience, set degradation weights for each component to determine the impact degree of different components on the health state of the ACM, and weighted sum the reconstruction errors of each component according to their degradation weights to construct a health indicator to comprehensively reflect the overall health status of the ACM and provide a basis for subsequent remaining life prediction; the expression is as follows: ; In the formula, is the total number of components, is the degradation weight of the th component, and the historical fault data includes the time, type, and scope of influence of the fault; is the th component's reconstruction error; Based on the LSTM neural network, an index prediction model is constructed. The model structure includes an input layer, an LSTM layer, a fully connected layer, and an output layer. The LSTM layer is used to handle the long-term dependence problem in time series data, and the fully connected layer is used to map the output of the LSTM layer to the prediction dimension. Arrange the historical health indicators data in chronological order, divide it into an index training set and an index validation set, and determine the length of the input sequence and the prediction step , and use the index training set to train the index prediction model. With the goal of minimizing the mean square error, the Adam optimization algorithm is used to update the model parameters. At the same time, during the training process, the performance of the model is evaluated through the validation set, and the model hyperparameters are adjusted to improve the prediction accuracy of the model. Input the most recent number of data points into the trained index prediction model, and output the predicted values for the next time steps. Through the prediction of the trend, potential faults of the system can be detected in advance to provide support for maintenance decisions. Using a survival analysis model, such as the Cox proportional hazards model, use historical failure data to estimate the parameters of the survival analysis model. With the goal of maximizing the likelihood function, the maximum likelihood estimation method is used to update the model parameters. Input the predicted values and key parameters output by the index prediction model into the trained survival analysis model, output the predicted failure probabilities of the ACM at different time points, and compare the predicted failure probabilities at different time points with the preset failure threshold. Once it is detected that the predicted failure probability at a certain time point exceeds the failure threshold, it is determined that the ACM is about to fail, and thus the remaining life of the ACM can be estimated.

[0031] Figure 6 Embodiment 2 Please refer to , another embodiment provided by the present invention: a data monitoring system, including: a data reading module, a diagnostic analysis module, a prediction module, and a feedback module. The data reading module is used to obtain offline data related to the aircraft air conditioner ACM, construct an ACM data set, preprocess the data in the ACM data set, set the feature engineering processing method, set a dynamic window, calculate the dynamic correlation coefficient of the compressor parameters and the turbine cooler parameters for coupled feature extraction, and at the same time, cluster the data according to the flight phase to divide the working conditions, and generate a structured time series data set. The diagnostic analysis module is used to obtain the baseline data of the ACM of the same model aircraft, set the baseline fault judgment method, judge the latest data in the structured time series dataset, combine the static threshold and parameter change rate analysis, screen out abnormal parameters, and combine the actual temperature adjustment effect to judge whether the ACM fails. It also sets the coupled fault location method to diagnose the faulty ACM, uses Granger causality analysis and decision tree algorithm to quantify the interactive influence between components, distinguish true faults from pseudo anomalies, accurately locate the faulty components, deeply analyze the cause of the fault, classify the fault according to the severity of the fault, and send corresponding maintenance instructions for different levels of faults to notify the maintenance personnel to perform maintenance; The prediction module is used to predict potential faults of the fault-free ACM through health index modeling and multi-model fusion, and judge the remaining life of the faulty components according to the fault prediction results; obtain the fault-free ACM, set the fault prediction method, calculate the predicted health index of the ACM to judge whether the ACM fails, and calculate the remaining life of the ACM; The feedback module is used to receive the feedback results after the staff perform the maintenance operation to judge the accuracy of fault diagnosis and prediction; for fault diagnosis, compare whether the actual maintenance result is consistent with the diagnosis result. If not, identify the specific difference problem and make targeted updates and modifications; for fault prediction, compare the predicted fault time or flight with the actual situation. If the predicted fault time or flight is quite different from the actual situation, update the model parameters, improve the algorithm or add new influencing factor variables to improve the accuracy of fault diagnosis and prediction, continuously improve the performance of monitoring the ACM, enhance the adaptability to complex fault modes, ensure the accuracy and robustness of the monitoring system, and form a closed-loop improvement mechanism.

[0032] Furthermore, the baseline fault judgment method includes: Through the flight parameter card, collect the complete operation data of at least fault-free flights in the initial stage of the same model aircraft leaving the factory, and based on the judgment rules in different flight stages, screen out the data in the cruise stage to generate a baseline database; For the key parameters in the baseline database, calculate the mean and standard deviation, and use the 3σ principle to construct a parameter baseline interval for the key parameters; Integrate the mean and baseline interval of each key parameter to construct a standard ACM performance model; Read the latest data in the structured time series dataset, input it into the standard ACM performance model for preliminary anomaly judgment, and construct an abnormal parameter set and a normal parameter set according to the output results; Read the normal parameters in the normal parameter set. For each normal parameter, select the latest continuous A sliding window, within each sliding window, the derivatives between adjacent data points within the window are calculated in sequence to reflect the rate of change of the parameter, obtaining a derivative sequence within the window; For each sliding window, analyze the change of the derivative within the window, calculate the mean and variance of the derivative sequence, and compare the calculated derivative mean and variance with the pre-set normal fluctuation range data; If the mean of the derivative sequence is not within the normal mean range, or the variance of the derivative sequence exceeds the normal variance range, mark that there is an abnormal change in the derivative of the corresponding normal parameter within the window; otherwise, there is no abnormality in the normal parameter; For continuous sliding windows, if there are abnormal changes in the derivatives within continuous sliding windows, it is determined that there is an abnormality in the corresponding normal parameter, and the corresponding normal parameter is updated to an abnormal parameter and migrated to the abnormal parameter set; otherwise, there is no abnormality in the corresponding normal parameter; Obtain the actual temperature adjustment effect in the aircraft and the expected temperature adjustment effect , and calculate the temperature difference ; Define the key parameters in the abnormal parameter set as abnormal parameters, obtain the number of abnormal parameters , set the temperature difference threshold to , the abnormal threshold to , and determine whether the ACM has a fault; When or , it is determined that the ACM has a fault; when and , it is determined that the ACM has no fault.

[0033] Furthermore, the coupled fault location method includes: When the ACM has a fault, in the structured time series dataset, according to the timestamp information, extract all data segments corresponding to the latest timestamp; Use the multi-variable time series association rule mining method, such as Granger causality analysis, to construct a bivariate regression model, evaluate the predictive ability of the past value of one variable on the current value of another variable, and introduce the Akaike information criterion (AIC) to determine the optimal lag order of the model. The AIC criterion can find a balance between the model complexity and the goodness of fit, ensure the accuracy of the model, and use the AIC criterion to determine the lag order , for different lag orders calculate the AIC value, and select the that makes the AIC value the smallest as the optimal lag order; Estimate the parameters of the bivariate regression model using the least squares method, calculate the residual sum of squares RSS, and minimize the residual sum of squares to accurately fit the data with the bivariate regression model and determine the parameters to be estimated; To quantify the degree of mutual influence between the compressor and turbine parameters, calculate the causal relationship strength ; Conduct a Granger causality test to determine whether there is a causal association between the parameters of the compressor and the turbine cooler; if , indicating that the lagged values of the turbine cooler variable contribute significantly to the prediction of the current value of the compressor variable, then it is determined that the turbine cooler variable is the Granger cause of the compressor variable; if , indicating that the lagged values of the compressor variable contribute significantly to the prediction of the current value of the turbine cooler variable, then it is determined that the compressor variable is the Granger cause of the turbine cooler variable; Based on the verification result of the Granger causality, make a preliminary judgment on the fault type to determine whether there is a Granger causal relationship between the compressor and the turbine cooler; If there is no causal relationship, indicating that there is no obvious mutual influence between the compressor and turbine cooler parameters, and the fault occurs independently or is caused by other non-coupling factors, then determine the fault type as a false anomaly and perform false anomaly processing; If there is a causal relationship, judge whether the dynamic correlation coefficient exceeds the preset normal correlation range; if , indicating that the dynamic correlation coefficient exceeds the normal correlation range and is marked as an abnormal correlation; if , indicating that the dynamic correlation coefficient is within the normal correlation range, determine the fault type as a false anomaly and perform false anomaly processing; where is the Z-score value of the correlation coefficient within the sliding window, is the normal correlation threshold; When it is determined that the dynamic correlation coefficient is an abnormal correlation, combine the flight phase information to judge the fault type; if the aircraft is in the cruise phase, determine the fault type as a true fault and perform true fault processing, and determine the faulty component and cause by matching historical maintenance cases; if the aircraft is not in the cruise phase, considering the complex operating environment of the aircraft during takeoff and landing and frequent parameter fluctuations, the current anomaly is caused by transient interference, determine the fault type as a false anomaly and perform false anomaly processing; According to the preset fault classification rules, classify the fault handling results and generate a maintenance report form, including the fault location report, level, and type; where the fault handling includes true fault handling and false anomaly handling, and the output of the fault handling is the fault location report.

[0034] Furthermore, the fault prediction method includes: When the ACM has no faults, construct a deep autoencoder model; Based on the structured time series dataset, divide it into a training set and a test set, and use the training set to train the deep autoencoder model to minimize the reconstruction error as the goal, and use the stochastic gradient descent algorithm to update the model parameters; Use the trained deep autoencoder model to reconstruct the data in the test set, calculate the reconstruction error, to evaluate the data reconstruction ability of the deep autoencoder model and reflect the degree of data abnormality; Based on historical fault data and expert experience, set degradation weights for each component to determine the impact degree of different components on the health state of the ACM, and perform weighted summation on the reconstruction errors of each component according to their degradation weights to construct a health index to comprehensively reflect the overall health status of the ACM and provide a basis for subsequent remaining life prediction; Based on the LSTM neural network, construct an index prediction model. The model structure includes an input layer, an LSTM layer, a fully connected layer, and an output layer; the LSTM layer is used to handle the long-term dependence problem in time series data, and the fully connected layer is used to map the output of the LSTM layer to the prediction dimension; Arrange the historical health index data in chronological order, divide it into an index training set and an index validation set, and determine the length of the input sequence and the prediction step and use the index training set to train the index prediction model. With the goal of minimizing the mean square error, use the Adam optimization algorithm to update the model parameters. At the same time, during the training process, evaluate the model performance through the validation set and adjust the model hyperparameters to improve the prediction accuracy of the model; Input the latest number of data points into the trained index prediction model, and output the prediction values for the next number of time steps Through the prediction of the trend, discover potential faults in the system in advance and provide support for maintenance decisions; Use a survival analysis model, such as the Cox proportional hazards model, and use historical fault data to estimate the parameters of the survival analysis model. With the goal of maximizing the likelihood function, use the maximum likelihood estimation method to update the model parameters; The output of the index prediction model The predicted values and key parameters are input into the trained survival analysis model to output the predicted failure probabilities of the ACM at different time points, and the predicted failure probabilities at different time points are compared with the preset failure threshold. Once it is detected that the predicted failure probability at a certain time point exceeds the failure threshold, it is determined that the ACM is about to fail, thereby estimating the remaining life of the ACM.

[0035] In summary of the above embodiments, the present invention extracts offline data, flight parameters and maintenance records by connecting the flight parameter card, generates an ACM data set in combination with environmental data, uses a dynamic window to extract the dynamic correlation coefficients of compressor and turbine parameters, clusters and standardizes them according to flight phases to generate structured time series data; constructs a benchmark database, screens abnormal parameters through static threshold detection and dynamic derivative analysis, combines temperature difference adjustment verification to judge faults, uses Granger causality analysis and decision trees to locate coupling faults, distinguishes true faults from pseudo anomalies, matches true faults with historical cases, and checks sensor noise for pseudo anomalies; constructs a deep autoencoder to calculate the reconstruction error, generates a health index in combination with degradation weights. LSTM predicts future trends, and a survival analysis model estimates the remaining life; verifies the accuracy of diagnosis and prediction according to the maintenance results, updates model parameters or algorithms, and continuously improves the system performance.

[0036] The above are only the preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the concept of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. ACM data analysis method, characterized in that, Including: Construct an ACM dataset, set the feature engineering processing method, extract coupled features by calculating the dynamic correlation coefficient through setting a dynamic window, and divide the working conditions to generate a structured time series dataset; Obtain the ACM baseline data of the same model aircraft, set the baseline fault judgment method, judge the latest data in the structured time series dataset, screen abnormal parameters, and combine with the actual temperature adjustment effect to judge whether the ACM is faulty. Set the coupled fault location method to diagnose the faulty ACM, distinguish true faults from pseudo anomalies, and at the same time set the prediction scheduling method to predict faults for the fault-free ACM and conduct health status prediction and life assessment.

2. The ACM data analysis method according to claim 1, wherein: When generating the structured time series dataset, the feature engineering processing method includes: Read the timestamps of the ACM dataset, calculate the difference between adjacent timestamps, and define the mode of the differences as the time resolution; Conduct a response test on the ACM, simulate the operation under different working conditions, monitor the fluctuation characteristics of key parameters, and determine the time range when the key parameters exceed the preset change range; Dynamically adjust the window size and sliding step based on a preset proportionality coefficient; Slide the window along the time series, perform Pearson correlation analysis on the compressor parameters and turbine cooler parameters within each window to generate a dynamic correlation coefficient; Associate the dynamic correlation coefficient with the start timestamp of the corresponding window to form a coupled feature time series.

3. The ACM data analysis method according to claim 2, characterized in that, The feature engineering processing method further includes: Divide the climb, cruise, and descent stages according to flight parameters, traverse the ACM dataset, and mark the data at each time point with the flight stage based on the preset flight stage judgment rules; Based on the recognition results of the flight stage and the dynamic correlation coefficient, save the data in the ACM dataset and the coupled feature time series to three subsets of climb, cruise, and descent respectively; Perform standardization processing on the data of each subset; Take the timestamp as the row index, arrange the key parameters and the dynamic correlation coefficient as a two-dimensional matrix to generate a structured time series dataset.

4. The ACM data analysis method according to claim 3, wherein: When judging whether the ACM is faulty, the baseline fault judgment method includes: Obtain the fault-free operation data of the same model aircraft in the initial stage of factory production, screen the cruise stage data based on the flight stage judgment rules, and generate a baseline database; Calculate the mean and standard deviation of the key parameters in the baseline database, and construct a parameter baseline interval using the 3σ principle; Integrate the mean of each key parameter and the parameter baseline interval to construct a standard ACM performance model; Read the latest data in the structured time series dataset, input it into the standard ACM performance model, compare it with the parameter baseline interval, identify abnormal parameters, and construct an abnormal parameter set and a normal parameter set.

5. The ACM data analysis method according to claim 4, wherein The baseline fault judgment method further includes: Read the normal parameters in the normal parameter set, calculate the derivatives between adjacent data points in the continuously sliding window in sequence, and obtain the derivative sequence within the window; For any single sliding window, calculate the mean and variance of the derivative sequence for dynamic trend detection; If the mean of the derivative sequence is not within the preset normal mean range or the variance exceeds the preset normal variance range, mark that there is an abnormal derivative change of the normal parameter within the sliding window; otherwise, there is no abnormal derivative change of the normal parameter; Perform anomaly judgment on consecutive sliding windows; if there are abnormal derivative changes in all consecutive sliding windows, update the normal parameter to an abnormal parameter and transfer it to the abnormal parameter set; otherwise, the normal parameter has no anomaly; The actual temperature adjustment effect in the computer cabin is measured, and the temperature difference is calculated , and a comprehensive judgment is made on whether there is a fault in the ACM; When or , it is determined that the ACM has a fault; when and , it is determined that the ACM has no fault; where is the number of abnormal parameters, is the temperature difference threshold, is the abnormal threshold.

6. The ACM data analysis method according to claim 5, wherein: When diagnosing a faulty ACM, the coupled fault location method includes: In the structured time series dataset, extract the data segment corresponding to the latest timestamp according to the timestamp information; Use the Granger causality analysis method to construct a bivariate regression model based on compressor parameters and turbine cooler parameters to evaluate the lag prediction ability between variables; Using the AIC criterion, calculate the AIC values for different lag orders and select the one that minimizes the AIC value as the optimal lag order; Estimate the parameters of the bivariate regression model based on the least squares method, calculate the sum of squared residuals and minimize the fitting data, and calculate the causal relationship strength between compressor and turbine cooler parameters; Judge whether there is a causal relationship between parameters through Granger causality test; if , then it is determined that the parameters of the turbine cooler are the Granger cause of the compressor parameters; if , then it is determined that the compressor parameters are the Granger cause of the turbine cooler parameters; where , are the parameters of the bivariate regression model.

7. The ACM data analysis method according to claim 6, wherein The coupled fault location method further includes: Based on the results of the Granger causality test, determine whether there is a causal relationship between compressor and turbine cooler parameters; If there is no causal relationship, determine the fault type as a pseudo anomaly and perform pseudo anomaly processing; If there is a causal relationship, determine whether the dynamic correlation coefficient exceeds the preset normal correlation range; If , determine that the fault type is a pseudo anomaly and perform pseudo anomaly processing; where is the Z-score value of the correlation coefficient within the sliding window, is the normal correlation threshold; If , determine the fault type in combination with the flight phase information; if the aircraft is in the cruise phase, determine that the fault type is a true fault and perform true fault handling; otherwise, determine that the fault type is a false anomaly and perform false anomaly handling; Generate a maintenance report according to the fault type.

8. The ACM data analysis method according to claim 7, characterized in that, The true fault handling includes: Based on the abnormal parameter set, obtain the components corresponding to the abnormal parameters in the abnormal parameter set to determine the candidate faulty components; Call the historical maintenance case library to screen the historical cases related to the candidate faulty components; Calculate the cosine similarity between the current fault data and the data of each historical case; Sort the historical cases according to the cosine similarity and output the top 3 fault causes and confidence levels with the highest similarity; Generate a fault location report including the faulty component, cause and confidence level, where the confidence level is the cosine similarity value.

9. The ACM data analysis method according to claim 8, wherein The pseudo anomaly processing includes: Extract sensor data from the structured time series dataset, check the data validity and continuity point by point, and calculate the continuity index; Compare the continuity index with the preset continuity threshold. If the continuity index is lower than the continuity threshold, mark the corresponding sensor as suspected faulty; otherwise, the sensor is normal; Perform noise level analysis on the sensor data, compare the calculated noise index with the preset noise threshold. If the noise index exceeds the noise threshold, mark the sensor as to be calibrated; otherwise, the sensor is normal; Construct a pseudo anomaly pattern library to store the characteristic data of historical pseudo anomaly events and manage them by classification; Use the dynamic time warping algorithm to match the sensor data without anomaly with the patterns in the pseudo anomaly pattern library and calculate the matching distance; If the matching distance is less than the preset distance threshold, output the corresponding pseudo anomaly type; Otherwise, update the pseudo - anomaly to a true fault and handle the true fault; Generate a fault location report including sensor status and pseudo - anomaly type.

10. The ACM data analysis method according to claim 9, wherein The method further includes using a fault prediction method to determine whether the ACM fails; The fault prediction method includes: Construct a deep auto - encoder model, train and evaluate the deep auto - encoder model using the structured time - series dataset, and calculate the reconstruction error based on the output of the deep auto - encoder model; Based on historical fault data, set a degradation weight for each component, sum the reconstruction errors of each component weighted by their degradation weights, and construct a health index; Build an index prediction model, train the index prediction model with historical health index data, and predict the health index trend in future time steps, where the length of the input sequence is , the prediction step length is ; Input the most recent health indicator data points into the indicator prediction model, and output the predicted values of the health indicators for the next time steps; Construct a survival analysis model based on historical fault data, input the predicted value of the health index and key parameters into the survival analysis model, and calculate the predicted failure probability at different time points; Compare the predicted failure probabilities at different time points with a preset failure threshold. Once it is detected that the predicted failure probability at a certain time point exceeds the failure threshold, determine that the ACM is about to fail and output the remaining life of the ACM.

11. A data monitoring system for implementing the ACM data analysis method according to any one of claims 1-10, characterized in that, It includes: A data reading module, a diagnostic analysis module, a prediction module, and a feedback module; The data reading module is used to construct an ACM dataset, set a dynamic window, extract coupled features based on the dynamic correlation coefficient, divide working conditions, and generate a structured time - series dataset; The diagnostic analysis module is used to screen out abnormal parameters by combining static thresholds and parameter change rates, and judge whether the ACM is faulty in combination with the actual temperature adjustment effect, and diagnose the faulty ACM to distinguish true faults from pseudo - anomalies; The prediction module is used to predict potential faults of the non - faulty ACM, determine whether the ACM fails according to the fault prediction result, and calculate the remaining life of the ACM; The feedback module is used to receive the feedback result after the staff performs maintenance operations to judge the accuracy of fault diagnosis and prediction, and perform update operations.

Images