An aero-engine parameter-oriented comprehensive early warning system and method

Abstract

The application discloses an aero-engine parameter-oriented comprehensive early warning system and method, relates to the technical field of aero-engine early warning, and comprises the following steps: acquiring multiple-channel operation parameters of an aero-engine, performing uniform time reference resampling, dividing a sliding window, establishing a stage environment bin identification, calculating a sliding window representative quantity of a consistent parameter pair, and establishing a fleet reference table; calculating a logarithmic ratio residual, generating a sliding window residual list, calculating a robust deviation quantity, and generating a local sliding window evidence record; acquiring a local comprehensive evidence strength according to the local sliding window evidence record, calculating a persistence index, and generating a single-machine decision record; gathering a same-bin fleet snapshot table, calculating a mechanism commonality proportion, generating a cooperative conclusion record, and acquiring a batch disposal list. The application improves the accuracy and environmental adaptability of parameter anomaly detection and improves the fine management level and risk prevention and control capability of aero-engine groups.

Description

Technical Field

[0001] This invention relates to the field of early warning technology for aero-engines, and in particular to a comprehensive early warning system and method for aero-engine parameters. Background Technology

[0002] In the field of aero-engine early warning, health assessment and early risk identification based on multi-channel operating parameters are important directions for aero-engine status monitoring. Conventional methods acquire multi-dimensional operating parameters and combine them with statistical analysis and signal feature extraction to comprehensively determine the operating status of aero-engines. By using methods such as sliding window processing, time series filtering, and quantile statistics to establish typical operating condition references, current observations are compared with historical baselines to identify changes in aero-engine operating trends, providing data support and reference for aero-engine operation and maintenance.

[0003] Existing early warning analysis methods based on multi-channel parameter comparison still have room for improvement. On the one hand, traditional algorithms often only perform time series feature analysis on a single engine, failing to fully reflect the group distribution characteristics of the same type of aircraft group under the same environmental conditions. On the other hand, in multi-engine collaborative scenarios, an effective quantitative characterization mechanism for the common aggregation law of abnormal mechanisms has not yet been established, and an effective mechanism for collaborative analysis of common mechanisms has not been formed, which restricts the rapid identification of potential risks at the aircraft group level and the accurate implementation of batch-based disposal. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a comprehensive early warning method for aero-engine parameters to solve the problems of existing technologies in quantifying the commonalities of mechanisms among aircraft groups and in achieving coordinated early warning and batch handling.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a comprehensive early warning method for aero-engine parameters, comprising,

[0008] Acquire multi-channel operating parameters of aero-engines, perform unified time base resampling, divide sliding windows, form stage labels and environment labels, establish stage environment bin identifiers, calculate the sliding window representative quantity of consistency parameter pairs, and establish a fleet reference table;

[0009] Based on the cluster reference table and the phase environment bin identifier, calculate the logarithmic ratio residual, generate a sliding window residual list, calculate the robust deviation, and generate the local sliding window evidence record;

[0010] Based on the evidence records in the local sliding window, obtain the comprehensive evidence strength of the local machine, calculate the persistence index, and generate a single-machine decision record;

[0011] Based on single-machine decision records, aggregate snapshot tables of the same-divided container group, calculate the common proportions of computer processing, generate collaborative conclusion records, and obtain batch disposal lists;

[0012] The calculation persistence index is expressed as follows:

[0013] ;

[0014] in, Indicates a persistent indicator. Indicates the number of consecutive sliding windows. This represents the time sliding window index variable. Indicates a sliding window The strength of the internal integrated evidence, Indicates the environmental compartment labeling in the phase. Lower quantile position The cluster residual quantile reference, Represents the positive part function;

[0015] The common proportion of the aforementioned mechanisms is expressed as follows:

[0016] ;

[0017] in, Indicates the category of responsibility mechanism Common proportions of mechanisms This indicates the total number of records in the abnormal single-machine decision record set. Indicates the index number of the abnormal single-machine decision record. Indicates the first The responsibility mechanism label for each abnormal single-machine decision record. Indicates the category of liability mechanism. This indicates an indicator function.

[0018] As a preferred embodiment of the comprehensive early warning method for aero-engine parameters described in this invention, the steps of acquiring multi-channel operating parameters of the aero-engine, performing unified time base resampling, dividing the sliding window, forming stage labels and environment labels, and establishing stage-environment bin identifiers are as follows:

[0019] The multi-channel operating parameters of the aero-engine are interpolated according to a unified time base to generate a multi-channel operating parameter sequence with a unified time base. Sliding windows are divided, effective sliding windows are obtained, and channel robust representative quantities are calculated.

[0020] Based on the robust representative quantity of the effective sliding window channel, combined with external environmental parameters, stage labels and environmental labels are formed, and stage environment bin identification is established.

[0021] Specifically, the formation stage label is obtained by reading the robust representative values ​​of fuel flow, engine speed, exhaust temperature, and vibration amplitude of each sliding window from the effective sliding window set. Within each sliding window, the rate of change of fuel flow, engine speed, and exhaust temperature are calculated by differentiating the robust representative values ​​at the two ends of the sliding window or by local linear fitting. The stage of the aero-engine is then determined, and the stage label is written into the sliding window record, marking the generation time and rule version number.

[0022] Specifically, to form an environmental label, discrete enumeration values ​​and corresponding codes are preset for altitude level, temperature status, pressure status, and humidity status. The altitude level value, temperature status value, pressure status value, and humidity status value within the same sliding window are concatenated in a fixed order of altitude level, temperature status, pressure status, and humidity status to form an environmental label.

[0023] As a preferred embodiment of the comprehensive early warning method for aero-engine parameters described in this invention, the calculation of the sliding window representative quantity of the consistency parameter pair and the establishment of the aircraft group reference table refer to calculating the sliding window representative quantity based on the stage environment bin identifier and the corresponding sliding window, summarizing the sliding window representative quantities of the same type of aero-engine under the same stage environment bin identifier, calculating the statistical characteristics of the aircraft group, and generating the aircraft group reference table.

[0024] As a preferred embodiment of the comprehensive early warning method for aero-engine parameters described in this invention, the step of calculating the logarithmic ratio residual based on the aircraft group reference table and the stage environment bin identifier, and generating the sliding window residual list, refers to determining the reference range of the sliding window according to the stage environment bin identifier based on the aircraft group reference table, and calculating the logarithmic ratio residual based on the sliding window representative quantity to generate the sliding window residual list.

[0025] As a preferred embodiment of the comprehensive early warning method for aero-engine parameters described in this invention, the calculation of robust deviation and generation of local sliding window evidence records refers to calculating robust deviation, obtaining responsibility pairs, and generating local sliding window evidence records based on the sliding window residual list.

[0026] As a preferred embodiment of the comprehensive early warning method for aero-engine parameters described in this invention, the specific steps for obtaining the comprehensive evidence strength of the local aircraft based on the local sliding window evidence record and calculating the persistence index are as follows:

[0027] Based on the local sliding window evidence records, construct a local sliding window evidence sequence arranged in chronological order, obtain the logarithmic ratio residuals of the historical sliding windows of the computer cluster, obtain the cluster residual quantile reference, and generate the comprehensive evidence strength of the local machine.

[0028] Based on the overall evidence strength of the machine and the phased environmental bin identification, combined with the residual quantile reference of the machine group, the persistence index is calculated.

[0029] As a preferred embodiment of the comprehensive early warning method for aero-engine parameters described in this invention, the generation of single-aircraft decision records refers to generating single-aircraft decision levels and obtaining single-aircraft decision records based on the comprehensive evidence strength, persistence index, and residual quantile reference of the aircraft group through logical condition rules.

[0030] As a preferred embodiment of the comprehensive early warning method for aero-engine parameters described in this invention, the steps of aggregating snapshot tables of aircraft groups based on single-aircraft decision records and calculating commonality ratios are as follows:

[0031] Using the sliding window number and the phase environment bin identifier as the joint search key, query all single-machine decision records and generate a snapshot table of the same bin group;

[0032] Based on the snapshot table of the same box-type machine group, abnormal single-machine decision records are filtered, an abnormal single-machine decision record set is constructed, and it is divided into abnormal subsets according to the responsibility mechanism label, and the commonality ratio of the calculation mechanism is calculated.

[0033] As a preferred embodiment of the comprehensive early warning method for aero-engine parameters described in this invention, the specific steps for generating collaborative conclusion records and obtaining batch disposal lists are as follows:

[0034] Based on the common proportion of mechanisms and the snapshot table of the same box group, the type of collaborative conclusion is determined by logical conditions, and a collaborative conclusion record is generated.

[0035] Based on the collaborative conclusion records and the snapshot table of the same sorting box group, generate a batch disposal list corresponding to the collaborative conclusion type and assign batch numbers.

[0036] Secondly, the present invention provides a comprehensive early warning system for aero-engine parameters, comprising,

[0037] The data binning module is used to acquire multi-channel operating parameters of aero-engines, perform unified time base resampling, divide sliding windows, form stage labels and environment labels, establish stage environment binning identifiers, calculate the sliding window representative quantity of consistency parameter pairs, and establish a fleet reference table.

[0038] The residual calculation module is used to calculate the logarithmic ratio residual based on the cluster reference table and the stage environment bin identifier, generate a sliding window residual list, calculate the robust deviation, and generate the local sliding window evidence record;

[0039] The stand-alone decision module is used to obtain the comprehensive evidence strength of the local machine based on the evidence records in the local sliding window, calculate the persistence index, and generate stand-alone decision records;

[0040] The cluster collaboration module is used to aggregate snapshot tables of clusters with the same container based on single-machine decision records, calculate the common proportions of data processing, generate collaborative conclusion records, and obtain batch disposal lists.

[0041] The beneficial effects of this invention are as follows: by establishing a phased environmental sub-binding identifier and a fleet reference table, an environmentally adaptive benchmark reference system is established, improving the accuracy of parameter anomaly detection and environmental adaptability; by aggregating the fleet snapshot table based on single-aircraft decision records and calculating the common proportions, collaborative identification and batch processing of risk patterns at the fleet level are achieved, forming a system from single-aircraft early warning to fleet collaborative decision-making, effectively improving the level of refinement and risk prevention and control capabilities of aero-engine group management. Attached Figure Description

[0042] To more clearly illustrate the technical solutions of the embodiments of the present invention, the 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.

[0043] Figure 1 This is a flowchart of a comprehensive early warning method for aero-engine parameters.

[0044] Figure 2 This is a schematic diagram of a comprehensive early warning system for aero-engine parameters.

[0045] Figure 3 A flowchart for generating a local sliding window evidence record.

[0046] Figure 4 A flowchart for generating a batch disposal list. Detailed Implementation

[0047] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0048] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0049] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0050] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a comprehensive early warning method for aero-engine parameters, comprising the following steps:

[0051] S1. Obtain multi-channel operating parameters of the aero-engine, perform unified time base resampling, divide the sliding window, form stage labels and environment labels, establish stage environment bin identifiers, calculate the sliding window representative quantity of consistency parameter pairs, and establish a fleet reference table.

[0052] The multi-channel operating parameters of the aero-engine are interpolated according to a unified time base to generate a multi-channel operating parameter sequence with a unified time base. Sliding windows are divided, effective sliding windows are obtained, and channel robust representative quantities are calculated.

[0053] Furthermore, a data access channel is established to receive multi-channel operating parameter sequences such as fuel flow, exhaust temperature, compressor inlet pressure, compressor outlet pressure, speed, vibration amplitude, flight altitude, ambient temperature, ambient air pressure, and ambient humidity. The channel identifier, original timestamp, unit, sampling method, and integrity flag are recorded. The start and end times and sampling frequency of each channel are obtained. A unified time reference clock is selected to generate a unified time grid covering the entire time period. The start point, step size, and total length of the unified time grid are marked in the metadata.

[0054] Time interpolation and deduplication are performed on each channel using a unified time grid. Specifically, multiple records at the same time point retain the first trigger record, missing test points are filled in according to the interpolation rules, and unusable intervals caused by link switching are marked as unusable. A multi-channel operating parameter sequence for resampling is generated.

[0055] Time alignment consistency verification is performed on the resampled multi-channel operating parameter time series. The time alignment consistency verification includes continuity check, monotonicity check, unit consistency check, and strict timestamp alignment check, generating a multi-channel operating parameter series with a unified time base.

[0056] Read the multi-channel operating parameter sequence with a unified time base, set the sliding window length and sliding step size according to the time axis length, take the start point of the time series as the start point of the first sliding window, slide along the time axis according to the sliding step size, generate multiple consecutive overlapping sliding windows in sequence, record the start time and end time of each sliding window, generate a list of sliding window start and end times, assign a sliding window number to each sliding window, and generate a sliding window index set.

[0057] Data integrity is checked channel by channel in each sliding window. If any channel is marked as unavailable within the sliding window, the entire sliding window is marked as invalid and removed, generating a set of valid sliding windows.

[0058] Within each effective sliding window, the median of each channel is calculated and used as a robust representative of the effective sliding window channels. The standard deviation of each channel is also calculated and used as a measure of the volatility of the effective sliding window channels.

[0059] Furthermore, a valid sliding window record is generated, which includes the sliding window number, the start and end time range of the sliding window, the robust representative quantity of each channel, the volatility measure of each channel, the integrity flag, and the generation time.

[0060] Based on the robust representative quantity of the effective sliding window channel, combined with external environmental parameters, stage labels and environmental labels are formed, and stage environment bin identification is established.

[0061] To generate stage labels, specifically, the robust representative values ​​of fuel flow, engine speed, exhaust temperature, and vibration amplitude of each sliding window are read from the set of effective sliding windows. Within each sliding window, the rate of change of fuel flow, engine speed, and exhaust temperature are calculated by differentiating the robust representative values ​​at the two ends of the sliding window or by local linear fitting. The stage of the aero-engine is then determined, and the stage label is written into the sliding window record, marking the generation time and rule version number.

[0062] Furthermore, when both the fuel flow rate change rate and the engine speed change rate are trending upwards, and the robust representative value of exhaust temperature continues to increase, the stage is labeled as the acceleration stage; when both the fuel flow rate change rate and the engine speed change rate are trending downwards, and the robust representative value of exhaust temperature is slowly decreasing or stabilizing, the stage is labeled as the deceleration stage; when the fuel flow rate change rate and the engine speed change rate are close to 0 with low fluctuations, the robust representative value of fuel flow is stable, and the standard deviation of the vibration amplitude is stable, the stage is labeled as the stable cruising stage; when the robust representative value of fuel flow gradually increases from 0, the robust representative value of engine speed changes from static to continuous increase, and the robust representative value of exhaust temperature suddenly increases, the stage is labeled as the starting stage; when the robust representative value of fuel flow gradually approaches 0, and the robust representative value of engine speed decreases significantly and tends to stop, the stage is labeled as the shutdown stage.

[0063] It should be noted that the rate of change of fuel flow and the rate of change of engine speed are almost zero. The fuel supply and engine speed of the aircraft engine remain in a steady-state equilibrium. During the steady-state cruise phase, the time derivatives of fuel flow and engine speed exhibit a zero mean or near-zero fluctuation after being averaged over multiple sliding windows.

[0064] Furthermore, robust representative quantities of external environmental parameters are read from the same sliding window record, including but not limited to robust representative quantities of flight altitude, ambient temperature, ambient air pressure, and ambient humidity.

[0065] Environmental labels are formed, specifically by determining the altitude level based on robust representative values ​​of flight altitude, including the ground zone, mid-altitude zone, and high-altitude zone; determining the temperature status based on robust representative values ​​of outside temperature, including hot, cold, and warm conditions; determining the pressure status based on robust representative values ​​of outside air pressure, including high pressure and low pressure; and determining the humidity status based on robust representative values ​​of outside humidity, including wet and dry conditions.

[0066] Furthermore, environmental labels are formed by logically combining the following parameters in the order of altitude, temperature, pressure, and humidity.

[0067] Specifically, discrete enumeration values ​​and corresponding codes are preset for altitude level, temperature status, pressure status, and humidity status. The altitude level value, temperature status value, pressure status value, and humidity status value within the same sliding window are concatenated in a fixed order of altitude level, temperature status, pressure status, and humidity status to form an environmental label.

[0068] Write the stage label and environment label into the corresponding sliding window record to generate a one-to-one correspondence between sliding window, stage and environment, and generate a set of labeled sliding windows.

[0069] Furthermore, the set of labeled sliding windows is read, and the stage label and environment label recorded in each sliding window are concatenated in a fixed order to generate a stage environment bin identifier. The stage environment bin identifier adopts an unambiguous string or enumeration encoding form.

[0070] Perform a consistency check on the sliding window records under the same stage environment bin identifier to ensure that the stage label and environment label of all sliding window records in the bin have the same meaning. If there is a label conflict, relabel the label or remove the conflicting record.

[0071] Furthermore, an index table is created for each bin, where the index items include the stage environment bin identifier, the sliding window number list, the sliding window time range list, and the generation time, generating a sliding window set with bin index and a bin index table.

[0072] Based on the stage environment bin identifier and the corresponding sliding window, calculate the sliding window representative quantity, summarize the sliding window representative quantities of the same type of aero-engine under the same stage environment bin identifier, and generate a cluster reference table based on the statistical characteristics of the computer cluster.

[0073] Furthermore, the bin index table and the sliding window set with bin index are read, and a set of consistent parameter pairs is constructed according to the physical coupling relationship. The set of consistent parameter pairs includes combustion parameter pairs, aerodynamic parameter pairs and mechanical parameter pairs. The combustion parameter pairs consist of fuel flow rate and exhaust temperature, the aerodynamic parameter pairs consist of compressor after pressure and compressor before pressure, and the mechanical parameter pairs consist of speed and vibration amplitude.

[0074] Furthermore, for each sliding window and each pair of consistency parameters, the ratio of the channel robust representative quantities of the consistency parameter pair is used as the sliding window representative quantity of the consistency parameter pair. The sliding window representative quantity includes the combustion consistency ratio, the aerodynamic consistency ratio, and the mechanical consistency ratio. Specifically, the ratio of the robust representative quantity of fuel flow rate to the robust representative quantity of exhaust temperature is used as the combustion consistency ratio, characterizing the matching relationship between combustion energy input and heat release; the ratio of the robust representative quantity of compressor outlet pressure to the robust representative quantity of compressor inlet pressure is used as the aerodynamic consistency ratio, characterizing the compressor's compression efficiency and aerodynamic balance characteristics; and the ratio of the robust representative quantity of rotational speed to the robust representative quantity of vibration amplitude is used as the mechanical consistency ratio, characterizing the operational stability of rotating components.

[0075] Furthermore, the system is grouped according to the stage environmental bin identifier and consistency parameter pairs, and the sliding window representative quantities of all consistency parameter pairs of the same type of aero-engine are aggregated to generate a statistical sample set at the aircraft group level. For each consistency parameter pair of the stage environmental bin identifier, the statistics of the median of the aircraft group ratio and several quantile positions of the aircraft group ratio are obtained.

[0076] Generate a cluster reference table. Specifically, establish a reference record for each stage environment bin identifier and each consistency parameter pair. The content includes, but is not limited to, the stage environment bin identifier, the consistency parameter pair identifier, the cluster ratio median, the statistics of several quantile positions of the cluster ratio, the generation time, and the version number.

[0077] It should be noted that the group ratio includes combustion consistency ratio, aerodynamic consistency ratio, and mechanical consistency ratio.

[0078] S2. Based on the cluster reference table and stage environment binning identifier, calculate the logarithmic ratio residual, generate a sliding window residual list, calculate the robust deviation, and generate the local sliding window evidence record.

[0079] Based on the cluster reference table, the reference range of the sliding window is determined according to the phase environment sub-bin identifier. Combined with the representative quantity of the sliding window, the logarithmic ratio residual is calculated, and a sliding window residual list is generated.

[0080] Furthermore, the cluster reference table is read, and the cluster reference entry corresponding to the stage and environment to which the current sliding window belongs is determined by using the stage environment bin identifier as an index. Each reference entry includes the cluster ratio median and statistical percentile position, as well as the sliding window representative quantity.

[0081] Furthermore, the logarithmic ratio residual of the consistency parameter pairs within the sliding window is calculated, reflecting the degree of deviation of the current sliding window representative quantity from the median of the group ratio.

[0082] The logarithmic ratio residual is expressed as:

[0083] ;

[0084] in, Represents a sliding window The logarithmic ratio of the residuals Represents a sliding window The sliding window represents a quantity. This indicates that the environmental bins are labeled for the same stage. The median of the group ratio below, Indicates the number of the sliding window. Indicates the first A sliding window, This indicates the stage of environmental compartmentation.

[0085] Furthermore, a positive log-ratio residual indicates that the window representativeness of the current sliding window is higher than the median of the cluster ratio, while a negative log-ratio residual indicates that the window representativeness of the current sliding window is lower than the median of the cluster ratio. The log-ratio residual for each pair of consistency parameters is calculated to generate a list of sliding window residuals.

[0086] Based on the sliding window residual list, calculate the robust deviation, obtain the responsibility pair, and generate the local sliding window evidence record.

[0087] Furthermore, based on the sliding window residual list, robust deviations are calculated. Specifically, the absolute value of the logarithmic ratio of the residuals of the consistency parameter pairs is used as the robust deviation to measure the strength of the deviation of the consistency parameter pairs within the sliding window. The robust deviations include combustion robust deviations, aerodynamic robust deviations, and mechanical robust deviations.

[0088] Furthermore, within the same sliding window, the magnitudes of robust deviations are compared, and the pair of consistency parameters with the largest robust deviations is identified as the responsibility pair.

[0089] Based on robust deviations and responsibility pairs, a local sliding window evidence record is established. This local sliding window evidence record is stored in a queryable database in the form of structured information records, making the entire process traceable from the sliding window number, stage environment bin identifier to robust deviations and responsibility pairs.

[0090] Specifically, a structured information record is created for each sliding window. The structured information record includes the sliding window number, the stage environment bin identifier, the source information of the sliding window representative quantity, the calculation source path of the log ratio residual, the three robust deviations, the name of the responsibility pair and the corresponding log ratio residual, the version number of the cluster reference table, and the start and end time range of the sliding window.

[0091] Structured information records are written into a queryable database, and a unique retrieval identifier is assigned to each structured information record.

[0092] S3. Based on the evidence records in the local sliding window, obtain the comprehensive evidence strength of the local machine, calculate the persistence index, and generate a single-machine decision record.

[0093] Based on the local sliding window evidence records, a local sliding window evidence sequence arranged in chronological order is constructed. The logarithmic ratio residuals of the historical sliding windows of the computer cluster are used to obtain the cluster residual quantile references and generate the comprehensive evidence strength of the local machine.

[0094] Furthermore, the database storing local sliding window evidence records is accessed. Based on the current sliding window number to be processed, the corresponding local sliding window evidence record is read. Based on the stage environment bin identifier and consistency parameter pair type, a query key for querying the cluster reference table is generated. The isomorphic data of the most recent L sliding windows are pre-fetched from the historical local sliding window evidence records in chronological order to generate a local sliding window evidence sequence arranged in chronological order.

[0095] Use the query key to access the cluster reference table, and extract the cluster historical sliding window representative quantity set under the corresponding stage environmental bin identifier from the cluster reference table. The cluster historical sliding window representative quantity set includes the cluster historical sliding window representative quantity and the cluster ratio median.

[0096] Furthermore, for each consistency parameter pair type, a cluster residual quantile reference is constructed. Specifically, for each historical sliding window in the representative set of the cluster's historical sliding window, the absolute value of the logarithmic ratio residual is calculated. The absolute values ​​of the logarithmic ratio residuals are distributed, and the cluster residual quantile reference is determined according to the empirical quantile method. Using the stage environment bin identifier, consistency parameter pair type, and quantile position as index keys, the corresponding cluster residual quantile reference and its generation time are stored in an independent reference storage table. The version number of the cluster residual quantile reference is consistent with the cluster reference table for data traceability.

[0097] The residual quantile reference of the aircraft group is represented as follows:

[0098] ;

[0099] in, Indicates the residual quantile reference of the aircraft group. Candidate quantiles representing the absolute values ​​of the log-ratio residuals. Mathematical operators that take the least upper bound. Represents probability operators, Showing the history of the fleet in a sliding window The logarithm of the absolute value of the residuals. This indicates a sliding window displaying the fleet's history. This represents the quantile position parameter, with a value range of (0.80, 0.95).

[0100] It should be noted that, specifically, to obtain the quantile position parameter, the following steps are taken: First, historical aircraft group sliding window data are collected within the environmental binning identifiers of the same stage. The absolute value of the log-ratio residual for each sliding window is calculated, and the absolute values ​​of the log-ratio residuals are arranged in ascending order to obtain the cumulative distribution of the absolute values ​​of the log-ratio residuals. Based on engineering experience and statistical analysis requirements, a probability position with a value range of (0.80, 0.95) is selected as the quantile position parameter. When the quantile position parameter is less than 0.80, the aircraft group residual quantile reference value is easily affected by occasional disturbances or random noise and fluctuates significantly. When the quantile position parameter is greater than 0.95, the cumulative distribution of the absolute value of the log-ratio residuals may mask the true abnormal deviation, which is not conducive to early warning. The value range of (0.80, 0.95) ensures the sensitivity of identifying abnormal changes in the operating status of aero-engines, improves the robustness of the aircraft group residual quantile reference, and the calculated aircraft group residual quantile reference can reflect the typical fluctuation range of normal operation and effectively identify potential abnormal trends.

[0101] Furthermore, within each sliding window of the continuous time series, the combustion robustness deviation, aerodynamic robustness deviation, and mechanical robustness deviation are read. The maximum value among the three is selected as the local comprehensive evidence strength, and the corresponding mechanism name is recorded as the responsible mechanism for the local comprehensive evidence strength. A corresponding responsible mechanism label is generated. The responsible mechanism label is used to identify the direct mechanism source causing the current local comprehensive evidence strength deviation. The value of the responsible mechanism label includes combustion mechanism, aerodynamic mechanism, and mechanical mechanism. Only one responsible mechanism label is generated for each sliding window.

[0102] Based on the overall evidence strength of the machine and the phased environmental bin identification, combined with the residual quantile reference of the machine group, the persistence index is calculated.

[0103] Furthermore, on a continuous time series, the calculation interval is the current sliding window and several adjacent sliding windows before the current sliding window. The local comprehensive evidence strength of each sliding window is read one by one and compared with the cluster residual quantile reference corresponding to the stage environment bin identifier to calculate the persistence index of the current sliding window.

[0104] Specifically, the persistence indicator is expressed as:

[0105] ;

[0106] in, Indicates a persistent indicator. Indicates the number of consecutive sliding windows. This represents the time sliding window index variable. Indicates a sliding window The strength of the internal integrated evidence, Indicates the environmental compartment labeling in the phase. Lower quantile position The cluster residual quantile reference, This represents the positive part of the function.

[0107] Based on the comprehensive evidence strength, persistence index, and residual quantile reference of the machine cluster, a single machine decision level is generated through logical condition rules, and the single machine decision record is obtained.

[0108] Specifically, based on the local comprehensive evidence strength, persistence index, and cluster residual quantile reference corresponding to the stage environment bin identifier in the current sliding window, logical condition rules are executed in a fixed order. Specifically, when the local comprehensive evidence strength is higher than the cluster residual quantile reference and the persistence index is greater than 0, the single-machine decision level is high-risk warning; when the local comprehensive evidence strength is less than or equal to the cluster residual quantile reference but the persistence index is greater than 0, the single-machine decision level is trend deviation; when the local comprehensive evidence strength is less than or equal to the cluster residual quantile reference and the persistence index is equal to 0, the single-machine decision level is normal monitoring. Each single-machine decision level includes, but is not limited to, the sliding window number, stage environment bin identifier, local comprehensive evidence strength, persistence index, responsibility mechanism label, responsibility pair, and cluster reference table version number, generating a structured single-machine decision record.

[0109] It should be noted that the persistence index characterizes the continuity of the deviation of the aero-engine operating parameters over time. The value of the persistence index is determined by the difference between the integrated evidence strength of the aircraft and the residual quantile reference of the aircraft group between adjacent sliding windows. Specifically, when the integrated evidence strength of the aircraft is consistently higher than the residual quantile reference of the aircraft group within adjacent sliding windows, it indicates that the deviation has a continuous characteristic over time, and the persistence index is positive. When the integrated evidence strength of the aircraft is lower than or equal to the residual quantile reference of the aircraft group, it indicates that the deviation has not formed a continuous trend, and the persistence index is 0.

[0110] Furthermore, using the sliding window number and the stage environment bin identifier as the primary key, the single-machine decision record is persistently stored and a one-to-one mapping is established with the corresponding local sliding window evidence record. Any single-machine decision record can be traced back to the source data, the cluster reference table version, and the calculation basis of the cluster residual quantile reference.

[0111] S4. Based on single-machine decision records, aggregate snapshot tables of the same-divided container group, calculate common proportions, generate collaborative conclusion records, and obtain batch disposal lists.

[0112] Using the sliding window number and the phase environment sub-bin identifier as the joint search key, query all single-machine decision records and generate a snapshot table of the same sub-bin cluster.

[0113] Furthermore, using the sliding window number and the stage environment sub-bin identifier as the joint retrieval key, the database storing individual engine decision records is used to query the individual engine decision records of all aero engines under the same sliding window number and the same stage environment sub-bin identifier, and a snapshot table of the same sub-bin aircraft group is constructed.

[0114] It should be noted that in the same-packet aircraft group snapshot table, each row of the same-packet aircraft group snapshot table records the single-engine decision record of the corresponding aircraft engine under the joint search key, including but not limited to the aircraft engine identifier, sliding window number, stage environment packing identifier, comprehensive evidence strength, persistence index, responsibility mechanism label, responsibility pair, single-engine decision level, and version number of the aircraft group reference table.

[0115] Furthermore, in the summary area of ​​the same-bin aircraft cluster snapshot table, the number of aircraft engines participating in the construction of this same-bin aircraft cluster snapshot table is counted as the total number of aircraft engines in this cluster analysis, and the sliding window number and stage environment bin identifier corresponding to this same-bin aircraft cluster snapshot table are recorded.

[0116] Based on the snapshot table of the same box-type machine group, abnormal single-machine decision records are filtered, an abnormal single-machine decision record set is constructed, and it is divided into abnormal subsets according to the responsibility mechanism label, and the commonality ratio of the calculation mechanism is calculated.

[0117] Furthermore, the commonality ratio of mechanisms characterizes the degree of common risk at the machine group level. The commonality ratio of mechanisms allows for a unified comparison of the common contributions of combustion mechanisms, aerodynamic mechanisms, and mechanical mechanisms within the machine group, facilitating collaborative decision-making at the machine group level.

[0118] Specifically, in the snapshot table of the same container group, the individual machine decision records with decision levels of high risk warning and trend deviation are selected to construct an abnormal individual machine decision record set. The responsibility mechanism label of each individual machine decision record is retained in the abnormal individual machine decision record set. If the abnormal individual machine decision record set is empty, the commonality ratio of the recorded mechanism is 0 in the summary area of ​​the snapshot table of the same container group, and it directly enters the subsequent collaborative conclusion generation step. The collaborative conclusion is that the group is under normal collaborative monitoring.

[0119] When the abnormal single-machine decision record set is not empty, the abnormal single-machine decision record set is divided into combustion mechanism abnormal subset, aerodynamic mechanism abnormal subset and mechanical mechanism abnormal subset according to the responsibility mechanism label. The number of abnormal subset records in the three types of abnormal subsets and the total number of abnormal single-machine decision record sets are counted respectively. The commonality ratio of mechanisms is calculated to characterize the degree of commonality of each mechanism in the abnormality.

[0120] Specifically, the proportion of common mechanisms is expressed as follows:

[0121] ;

[0122] in, Indicates the category of responsibility mechanism Common proportions of mechanisms This indicates the total number of records in the abnormal single-machine decision record set. Indicates the index number of the abnormal single-machine decision record. Indicates the first The responsibility mechanism label for each abnormal single-machine decision record. Indicates the category of liability mechanism. Indicates an indicator function.

[0123] The common mechanism proportions include the common proportions of combustion mechanism, aerodynamic mechanism, and mechanical mechanism. The mechanism corresponding to the largest common mechanism proportion is taken as the dominant common mechanism source. The three types of common mechanism proportions and the dominant common mechanism source are recorded in the summary area of ​​the snapshot table of the same compartment group.

[0124] Based on the common proportion of mechanisms and the snapshot table of the same box group, the type of collaborative conclusion is determined by logical conditions, and a collaborative conclusion record is generated.

[0125] Based on the common proportion of mechanisms and individual aircraft decision records, collaborative conclusion records are generated. Specifically, the summary area of ​​the same container group snapshot table is read to obtain the total number of aircraft engines, the total number of abnormal individual aircraft decision record sets, the common proportion of the three types of mechanisms, and the common source of the dominant mechanism in this aircraft group analysis. From the same container group snapshot table, the record sets with the individual aircraft decision level of high risk warning and the record sets with the individual aircraft decision level of trend deviation are selected, and the responsibility mechanism label of each individual aircraft decision record is retained.

[0126] Furthermore, in the high-risk warning record set, the number of single-machine decision records whose responsibility mechanism label equals the common source of the dominant mechanism is counted; if the number of single-machine decision records whose responsibility mechanism label equals the common source of the dominant mechanism in the high-risk warning record set is not 0, and the total number of abnormal single-machine decision record sets is greater than 1, then the collaborative conclusion type under the current sliding window and stage environment bin identifier is determined as a cluster collaborative high-risk warning.

[0127] It should be noted that by statistically analyzing the number of single-aircraft decision records in the high-risk warning record set whose responsibility mechanism label is consistent with the common source of the dominant mechanism, it is determined whether there is a risk clustering phenomenon dominated by the same mechanism within the aircraft group. The minimum condition for the existence of a common mechanism is that the number of single-aircraft decision records in the high-risk warning record set whose responsibility mechanism label is equal to the common source of the dominant mechanism is not zero. This indicates that at least one aircraft engine has abnormal behavior consistent with the common source of the dominant mechanism under the current environmental binning identification, indicating that a risk response based on the same mechanism has appeared within the aircraft group.

[0128] Determining whether the total number of abnormal single-aircraft decision record sets is greater than 1 is the minimum condition for establishing a group-level collaborative judgment. Confirming whether the abnormal behavior in the group has multi-aircraft collaboration indicates that at least two or more aero engines have deviated under the same stage environmental sub-box identifier, confirming that the deviation is not a single-aircraft exception, but a collaborative abnormality within the group.

[0129] Furthermore, when the high-risk early warning conditions for swarm collaboration are not met, the number of single-machine decision records in the trend deviation record set where the responsibility mechanism label is equal to the common source of the dominant mechanism is counted. If the number of single-machine decision records in the trend deviation record set where the responsibility mechanism label is equal to the common source of the dominant mechanism is not 0, and the total number of abnormal single-machine decision record sets is greater than 1, then the collaboration conclusion type is determined to be swarm collaboration trend deviation.

[0130] It should be noted that the number of single-engine decision records in the statistical trend deviation record set whose responsibility mechanism label is consistent with the common source of the dominant mechanism is used to determine whether there is a deviation trend under the same mechanism in the fleet. When the number of single-engine decision records in the trend deviation record set whose responsibility mechanism label is equal to the common source of the dominant mechanism is not 0, it means that at least one aero-engine has a deviation mechanism consistent with the common source of the dominant mechanism under the current stage environmental binning identifier. The influence of the common source of the dominant mechanism has appeared in the fleet and is observable. This is the minimum condition for the existence of a common trend of mechanism.

[0131] To determine whether the total number of records in the abnormal single-engine decision record set is greater than 1, if only one aero-engine deviates, even if the deviation mechanism and the dominant mechanism have the same common source, it is only a single-engine individual response and is not enough to prove that a group coordination trend has formed. When the total number of records in the abnormal single-engine decision record set is greater than 1, it is the minimum condition for the establishment of a group coordination trend deviation, indicating that at least two or more aero-engines have a deviation trend dominated by the same mechanism under the same environmental bin identification at the same stage, indicating that the deviation trend has a group nature.

[0132] Furthermore, when neither the high-risk early warning condition for swarm collaboration nor the trend deviation condition for swarm collaboration is met, the collaboration conclusion type is determined to be normal swarm collaboration monitoring. This means that under the current stage environmental bin identification and sliding window number, the number of abnormal single-machine decision records is 0 or the abnormal single-machine decision records are scattered among multiple mechanisms and do not form a common cluster, and the swarm as a whole can maintain a normal monitoring state.

[0133] Establish collaborative conclusion records for the current sliding window number and the stage environment bin identifier. The collaborative conclusion records include, but are not limited to, the sliding window number, the stage environment bin identifier, the total number of aero engines analyzed in this fleet analysis, the total number of abnormal single-aircraft decision record sets, the common proportion of combustion mechanisms, the common proportion of aerodynamic mechanisms, the common proportion of mechanical mechanisms, the common source of the dominant mechanism, the type of collaborative conclusion, and the time of generation of collaborative conclusion. In the database, establish a one-to-one mapping between the collaborative conclusion records and the fleet snapshot table of the same bin. Any collaborative conclusion can be traced back to all single-aircraft decision records and sliding window evidence records involved in the judgment.

[0134] Based on the collaborative conclusion records and the snapshot table of the same sorting box group, generate a batch disposal list corresponding to the collaborative conclusion type and assign batch numbers.

[0135] Furthermore, at the operational decision-making level, a batch disposal list that can be directly executed is generated. Specifically, when the collaborative conclusion type is a high-risk warning for aircraft group collaboration, the aero-engine records with a single-aircraft decision level of high-risk warning and a responsibility mechanism label equal to the common source of the dominant mechanism are selected from the snapshot table of the same group of aircraft. The corresponding aero-engine identifier, sliding window number, stage environment sub-bin identifier, comprehensive evidence strength, persistence index, responsibility mechanism label, and responsibility pair are organized into batch disposal items. By summarizing all batch disposal items in the same batch, a high-risk batch disposal list corresponding to the sliding window and stage environment sub-bin identifier is generated.

[0136] When the collaborative conclusion type is a trend deviation of the aircraft group, select the aircraft engine records from the same container group snapshot table that have a single aircraft decision level of trend deviation and whose responsibility mechanism label is equal to the common source of the dominant mechanism. Compile the corresponding aircraft engine identification, comprehensive evidence strength, persistence indicators, responsibility mechanism label and responsibility pair into a trend deviation batch disposal list, which is used to formulate an encrypted monitoring plan or shorten the health assessment cycle.

[0137] When the collaborative conclusion type is normal monitoring of the aircraft group, no batch disposal entry for specific aero engines is generated. Only an empty batch disposal list is recorded to maintain the association with the collaborative conclusion record, indicating that the aircraft group does not need additional batch disposal actions under the phase environment bin identification.

[0138] Assign a batch number to each batch disposal list, establish a one-to-one mapping between the batch number and the collaborative conclusion record number, and record the stage environmental sub-binding identifier, sliding window number range, list of involved aero-engine identifiers, responsibility mechanism label, responsibility pair description, and generation time in the batch disposal list.

[0139] This embodiment also provides a comprehensive early warning system for aero-engine parameters, including:

[0140] The data binning module is used to acquire multi-channel operating parameters of aero-engines, perform unified time base resampling, divide sliding windows, form stage labels and environment labels, establish stage environment binning identifiers, calculate the sliding window representative quantity of consistency parameter pairs, and establish a fleet reference table.

[0141] The residual calculation module is used to calculate the logarithmic ratio residual based on the cluster reference table and the stage environment bin identifier, generate a sliding window residual list, calculate the robust deviation, and generate the local sliding window evidence record.

[0142] The stand-alone decision module is used to obtain the comprehensive evidence strength of the local machine based on the evidence records in the local sliding window, calculate the persistence index, and generate stand-alone decision records.

[0143] The cluster collaboration module is used to aggregate snapshot tables of clusters with the same container based on single-machine decision records, calculate the common proportions of data processing, generate collaborative conclusion records, and obtain batch disposal lists.

[0144] In summary, this invention establishes a phased environmental sub-binding identifier and a fleet reference table to create an environmentally adaptive benchmark reference system, thereby improving the accuracy of parameter anomaly detection and environmental adaptability. By aggregating fleet snapshot tables based on individual aircraft decision records and calculating commonality ratios, it achieves collaborative identification and batch processing of risk patterns at the fleet level, forming a system from individual aircraft early warning to fleet collaborative decision-making, effectively improving the level of precision management and risk prevention and control capabilities of aero-engine groups.

[0145] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A comprehensive early warning method for aero-engine parameters, characterized in that: include, Acquire multi-channel operating parameters of aero-engines, perform unified time base resampling, divide sliding windows, form stage labels and environment labels, establish stage environment bin identifiers, calculate the sliding window representative quantity of consistency parameter pairs, and establish a fleet reference table; Based on the cluster reference table and the phase environment bin identifier, calculate the logarithmic ratio residual, generate a sliding window residual list, calculate the robust deviation, and generate the local sliding window evidence record; Based on the evidence records in the local sliding window, obtain the comprehensive evidence strength of the local machine, calculate the persistence index, and generate a single-machine decision record; Based on single-machine decision records, aggregate snapshot tables of the same-divided container group, calculate the common proportions of computer processing, generate collaborative conclusion records, and obtain batch disposal lists; The calculation persistence index is expressed as follows: ； in, Indicates a persistent indicator. Indicates the number of consecutive sliding windows. This represents the time sliding window index variable. Indicates a sliding window The strength of the internal integrated evidence, Indicates the environmental compartment labeling in the phase. Lower quantile position The cluster residual quantile reference, Represents the positive part function; The common proportion of the aforementioned mechanisms is expressed as follows: ； in, Indicates the category of responsibility mechanism Common proportions of mechanisms This indicates the total number of records in the abnormal single-machine decision record set. Indicates the index number of the abnormal single-machine decision record. Indicates the first The responsibility mechanism label for each abnormal single-machine decision record. Indicates the category of liability mechanism. This indicates an indicator function.

2. The comprehensive early warning method for aero-engine parameters as described in claim 1, characterized in that: The specific steps for acquiring multi-channel operating parameters of the aero-engine, performing unified time base resampling, dividing the window into sliding windows, forming stage labels and environment labels, and establishing stage-environment bin identifiers are as follows: The multi-channel operating parameters of the aero-engine are interpolated according to a unified time base to generate a multi-channel operating parameter sequence with a unified time base. Sliding windows are divided, effective sliding windows are obtained, and channel robust representative quantities are calculated. Based on the robust representative quantity of the effective sliding window channel, combined with external environmental parameters, stage labels and environmental labels are formed, and stage environment bin identification is established. Specifically, the formation stage label is obtained by reading the robust representative values ​​of fuel flow, engine speed, exhaust temperature, and vibration amplitude of each sliding window from the effective sliding window set. Within each sliding window, the rate of change of fuel flow, engine speed, and exhaust temperature are calculated by differentiating the robust representative values ​​at the two ends of the sliding window or by local linear fitting. The stage of the aero-engine is then determined, and the stage label is written into the sliding window record, marking the generation time and rule version number. Specifically, to form an environmental label, discrete enumeration values ​​and corresponding codes are preset for altitude level, temperature status, pressure status, and humidity status. The altitude level value, temperature status value, pressure status value, and humidity status value within the same sliding window are concatenated in a fixed order of altitude level, temperature status, pressure status, and humidity status to form an environmental label.

3. The comprehensive early warning method for aero-engine parameters as described in claim 2, characterized in that: The calculation of the sliding window representative quantity of the consistency parameter pair and the establishment of the aircraft group reference table refer to calculating the sliding window representative quantity based on the stage environment bin identifier and the corresponding sliding window, summarizing the sliding window representative quantities of the same type of aero-engine under the same stage environment bin identifier, calculating the statistical characteristics of the aircraft group, and generating the aircraft group reference table.

4. The comprehensive early warning method for aero-engine parameters as described in claim 3, characterized in that: The process of calculating the logarithmic ratio residual and generating a sliding window residual list based on the cluster reference table and the stage environment bin identifier refers to determining the reference range of the sliding window according to the stage environment bin identifier based on the cluster reference table, and calculating the logarithmic ratio residual in combination with the representative quantity of the sliding window to generate a sliding window residual list.

5. The comprehensive early warning method for aero-engine parameters as described in claim 4, characterized in that: The calculation of robust deviation and generation of local sliding window evidence records refers to calculating robust deviation based on the sliding window residual list, obtaining the responsibility pair, and generating local sliding window evidence records.

6. The comprehensive early warning method for aero-engine parameters as described in claim 5, characterized in that: The specific steps for obtaining the comprehensive evidence strength and calculating the persistence index based on the local sliding window evidence records are as follows: Based on the local sliding window evidence records, construct a local sliding window evidence sequence arranged in chronological order, obtain the logarithmic ratio residuals of the historical sliding windows of the computer cluster, obtain the cluster residual quantile reference, and generate the comprehensive evidence strength of the local machine. Based on the overall evidence strength of the machine and the phased environmental bin identification, combined with the residual quantile reference of the machine group, the persistence index is calculated.

7. The comprehensive early warning method for aero-engine parameters as described in claim 6, characterized in that: The generation of single-machine decision records refers to generating single-machine decision levels and obtaining single-machine decision records based on the comprehensive evidence strength, persistence index, and cluster residual quantile reference of the local machine, through logical condition rules.

8. The comprehensive early warning method for aero-engine parameters as described in claim 7, characterized in that: The specific steps for aggregating snapshot tables of the same-class container group based on single-machine decision records and calculating common proportions by computer are as follows: Using the sliding window number and the phase environment bin identifier as the joint search key, query all single-machine decision records and generate a snapshot table of the same bin group; Based on the snapshot table of the same box-type machine group, abnormal single-machine decision records are filtered, an abnormal single-machine decision record set is constructed, and it is divided into abnormal subsets according to the responsibility mechanism label, and the commonality ratio of the calculation mechanism is calculated.

9. The comprehensive early warning method for aero-engine parameters as described in claim 8, characterized in that: The specific steps for generating collaborative conclusion records and obtaining batch disposal lists are as follows: Based on the common proportion of mechanisms and the snapshot table of the same box group, the type of collaborative conclusion is determined by logical conditions, and a collaborative conclusion record is generated. Based on the collaborative conclusion records and the snapshot table of the same sorting box group, generate a batch disposal list corresponding to the collaborative conclusion type and assign batch numbers.

10. A comprehensive early warning system for aero-engine parameters, based on the comprehensive early warning method for aero-engine parameters as described in any one of claims 1 to 9, characterized in that: include, The data binning module is used to acquire multi-channel operating parameters of aero-engines, perform unified time base resampling, divide sliding windows, form stage labels and environment labels, establish stage environment binning identifiers, calculate the sliding window representative quantity of consistency parameter pairs, and establish a fleet reference table. The residual calculation module is used to calculate the logarithmic ratio residual based on the cluster reference table and the stage environment bin identifier, generate a sliding window residual list, calculate the robust deviation, and generate the local sliding window evidence record; The stand-alone decision module is used to obtain the comprehensive evidence strength of the local machine based on the evidence records in the local sliding window, calculate the persistence index, and generate stand-alone decision records; The cluster collaboration module is used to aggregate snapshot tables of clusters with the same container based on single-machine decision records, calculate the common proportions of data processing, generate collaborative conclusion records, and obtain batch disposal lists.

Images