Icing detection method and system based on performance change of rotating part

Abstract

The invention discloses an icing detection method and system based on performance change of a rotating part, relates to the technical field of aviation, and aims to solve the problems that under cold and humid weather conditions, the rotating part of an aircraft and a wind driven generator is easy to freeze, which causes damage to the aerodynamic shape of the aircraft, reduction of lift force, deterioration of control performance and even causes flight accidents. The mass distribution and aerodynamic characteristics of the blades of the wind driven generator can be changed when the blades are frozen, and unbalanced vibration and power generation efficiency reduction are caused. In the prior art, an icing detector based on optics, capacitance and ultrasonic principles has the defects of load limitation, sensitive mounting position and the like, and is difficult to adapt to complex flight attitudes of small and medium-sized aircrafts and electric vertical take-off and landing aircrafts. The invention provides an icing detection method and system based on performance change of a rotating part, and aims to provide an icing detection solution which does not need additional equipment and is decoupled from a flight attitude so as to improve the running safety and efficiency of an aircraft.

Description

Technical Field

[0001] The present invention relates to the field of aircraft technology, and in particular to an ice detection method and system based on performance changes of rotating parts. Background Art

[0002] When an aircraft flies under cold and humid weather conditions, components such as wings, propellers, and engine air intakes are very susceptible to ice. Similarly, in the field of new energy, wind turbine blades operate in a low-temperature, high-humidity environment, and ice will also appear on the surface. Once these rotating parts are frozen, they will cause many serious problems. For example, ice on an aircraft will destroy the aerodynamic shape of the wing, increase flight resistance, reduce lift, deteriorate flight control performance, and even cause catastrophic flight accidents; ice on wind turbine blades will change their own mass distribution and aerodynamic characteristics, causing the fan to intensify unbalanced vibrations, significantly reduce power generation efficiency, and may also cause irreversible damage to the blade structure. Therefore, timely and accurate detection of icing conditions is of vital importance to ensuring the safe operation of related equipment and maintaining normal working efficiency.

[0003] Currently, existing ice detection technology solutions on the market, such as detectors based on optical principles, capacitance principles, and ultrasonic principles, can monitor ice phenomena to a certain extent, but they all have significant defects. On the one hand, in order to achieve the detection function, these traditional detectors all require the installation of special equipment on the aircraft. For small and medium-sized aircraft, the load margin is extremely limited. The weight of the additional equipment not only occupies the precious load quota, but also triggers a chain reaction in the overall design process of the aircraft, forcing the design team to rebalance multiple performance indicators such as power, endurance, and control, which greatly increases the complexity and difficulty of the design, and may even affect the basic performance parameters of the aircraft.

[0004] On the other hand, traditional detectors have almost stringent requirements for installation positions. During the flight of conventional fixed-wing aircraft, a suitable installation position may be barely found. Although the detection effect is restricted by certain environmental factors, it can still maintain basic operation. However, with the emergence of new aircraft such as electric vertical take-off and landing aircraft in the low-altitude economy, the situation has become increasingly complicated. During the vertical take-off and landing phase of such aircraft, the airflow field changes dramatically compared to the traditional flight posture. If traditional detectors still follow the conventional installation position layout, they will instantly fall into trouble: the optical path originally relied on for accurate detection is disturbed by turbulent airflow, and the refraction and reflection of light are abnormal; capacitive sensors are affected by complex and changeable electromagnetic fields and the sharp fluctuations in temperature and humidity caused by airflow, and the measurement stability collapses; ultrasonic detectors are difficult to accurately focus on the detection target under chaotic airflow vortices and changeable flight postures, resulting in a sharp increase in the risk of detection failure.

[0005] In summary, the existing ice detection technology cannot take into account the challenges brought by load limitations and flight attitude changes when dealing with small and medium-sized aircraft, especially electric vertical take-off and landing aircraft that meet the needs of low-altitude economic development, and it is difficult to accurately and efficiently complete the task of ice detection on rotating parts. Summary of the invention

[0006] In order to solve the problem that it is difficult for small and medium-sized aircraft with complex and changeable flight postures to accurately and efficiently complete the task of detecting ice on rotating parts, the present invention provides an ice detection method based on the performance change of rotating parts, which is characterized by comprising the following steps: S1. Input the flight parameters and rotating component parameters of the aircraft to be tested; S2. dynamically combining the flight parameters and the rotating component parameters by moving average, standardization and dimensionless conversion, storing the processed data in a database, and marking unmarked data as un-iced data by default; S3. When the total temperature of the incoming flow is ≥273.2K, it is directly determined as a non-icing condition without relying on other parameter analysis; S4. Using a machine learning model to identify abnormal disturbances, and filtering interference data by the change characteristics of the flight parameters and the rotating component parameters; S5. Based on the time series characteristics of the flight parameters and the rotating component parameters, a machine learning model is used to determine whether there is icing, the degree of icing and the speed; S6. Real-time feedback of recognition status through sound and light signals, allowing manual correction of marks, and maintaining the original judgment when no response is given; S7. storing the flight parameters, the rotating component parameters, the processed data and the identification status in a complete manner in a time series into a temporary storage medium; S8. Based on data integrity, all data in the temporary storage medium is divided into level 0 data, level 1 data, level 2 data and level 3 data; S9. Prioritize extracting the level 3 data, randomly balancing sample categories, and updating the machine learning model.

[0007] Further, in S1, the flight parameters include: wind speed, static temperature, static pressure, mean effective diameter of supercooled water droplets MVD and liquid water content LWC output by the wind tunnel measurement and control system; the rotating component parameters include: torque, rotation speed, inclination angle, and pitch angle.

[0008] Further, in S4, the abnormal disturbance phenomenon includes: turbulence, bird strike and gust.

[0009] Further, in S8, the level 0 data is invalid data or data that is duplicated with existing data, the level 1 data is low-quality data, the level 2 data is ordinary-quality data, and the level 3 data is high-quality data.

[0010] Furthermore, the updating of the machine learning model specifically includes: launching a classification learning operation, and updating and improving the machine learning model based on the prediction success rate and the reaction delay rate.

[0011] An ice detection system based on performance changes of rotating parts, the system comprising: a ground full-function system, an airborne detection system, a ground full-function system box body and an airborne detection system box body; The ground full-function system is located inside the ground full-function system box; The airborne detection system is located inside the airborne detection system box; The ground-based fully functional system is used to continuously optimize the machine learning model through deep analysis and data accumulation; The onboard detection system is used for rapid response, data filtering and implementation of local decisions; The ground full-function system includes: a data input interface, a data processing unit, an icing condition sensing unit, an icing sensing unit, an abnormal disturbance sensing unit, a temporary storage unit, a data output interface, a wireless transmission unit, a database unit, a machine learning unit, and an airborne system update unit; The data input interface is used to input the flight parameters and rotating component parameters of the aircraft to be tested; The data processing unit is used to perform dynamic combination processing on the flight parameters and the rotating component parameters by moving average, standardization and dimensionless conversion, store the processed data in a database, and mark the unmarked data as un-iced data by default; The icing condition sensing unit is used to directly determine that the inflow total temperature is ≥273.2K as a non-icing condition without relying on other parameter analysis; The abnormal disturbance sensing unit is used to identify abnormal disturbance phenomena using a machine learning model, and filter disturbance data according to the change characteristics of the flight parameters and the rotating component parameters; The icing sensing unit is used to determine whether icing, the degree of icing and the speed by using a machine learning model based on the time series characteristics of the flight parameters and the rotating component parameters; The data output interface and the wireless transmission unit are used to provide real-time feedback of the recognition status through sound and light signals, allowing manual correction of the mark, and maintaining the original judgment when there is no response; The temporary storage unit is used to store the flight parameters, the rotating component parameters, the processed data and the identification status in a temporary storage medium in full according to a time series; The database unit is used to classify all data in the temporary storage medium into level 0 data, level 1 data, level 2 data and level 3 data according to data integrity; The machine learning unit and the airborne system update unit are used to preferentially extract the level 3 data, randomly balance sample categories, and update the machine learning model.

[0012] Further, the airborne detection system includes: an airborne data input interface, an airborne data processing unit, an airborne icing condition sensing unit, an airborne icing sensing unit, an airborne abnormal disturbance sensing unit, an airborne temporary storage unit, an airborne data output interface, and an airborne wireless transmission unit; The airborne data input interface is used to input the flight parameters and rotating component parameters of the aircraft to be tested; The onboard data processing unit is used to perform dynamic combination processing on the flight parameters and the rotating component parameters by moving average, standardization and dimensionless conversion, store the processed data in a database, and mark the unmarked data as un-iced data by default; The airborne icing condition sensing unit is used to directly determine that the airflow is in a non-icing condition when the total temperature is ≥ 273.2K, without relying on other parameter analysis; The airborne abnormal disturbance perception unit is used to identify abnormal disturbance phenomena using a machine learning model and filter disturbance data according to the change characteristics of the flight parameters and the rotating component parameters; The airborne icing sensing unit is used to determine whether icing, the degree of icing and the speed by using a machine learning model based on the time series characteristics of the flight parameters and the rotating component parameters; The airborne data output interface and the airborne wireless transmission unit are used to provide real-time feedback of the recognition status through sound and light signals, allowing manual correction of the mark and maintaining the original judgment when there is no response; The onboard temporary storage unit is used to store the flight parameters, the rotating component parameters, the processed data and the identification status in a temporary storage medium in full according to a time series.

[0013] The beneficial effects of the present invention are: 1. The system is simple and highly compatible: no need to install special detection equipment, which reduces the workload of aircraft modification and avoids system compatibility issues. It is suitable for existing and new aircraft, shortens the R&D cycle and reduces costs.

[0014] 2. Accurate and reliable: Real-time monitoring of performance changes of rotating parts, and intelligent algorithms accurately analyze subtle parameter fluctuations caused by icing, providing pilots with a reliable basis for decision-making and improving flight safety.

[0015] 3. Energy saving and high efficiency: Utilizing existing system monitoring data, no additional energy supply is required, reducing the energy burden of the aircraft, extending the effective range, and conforming to the concept of green aviation.

[0016] 4. Adapt to complex scenarios: In complex weather and flight attitudes such as severe low-altitude convection and frequent changes in attitude of electric vertical take-off and landing aircraft, it can stably and continuously detect icing conditions, ensure safe take-off and landing under extreme conditions, expand the operational weather range, and enhance operational flexibility.

[0017] 5. Strong data support: Automatically collect, store and analyze icing data to provide big data support for aircraft design optimization, flight mission planning and maintenance strategy formulation, and promote the intelligent and refined development of the industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flow chart of an ice detection method based on performance changes of rotating parts; Figure 2 The schematic diagram is a method for detecting ice based on the performance change of rotating parts; Figure 3 This is the block diagram of the ground full-function system; Figure 4 This is the block diagram of the airborne detection system; Figure 5 This is a comparison chart between the actual icing conditions and the icing detection effect of the airborne detection system.

[0019] Reference numerals: In the figure, 1-ground full-function system, 2-airborne detection system, 101-ground full-function system box, 102-data input interface, 103-data processing unit, 104-icing condition perception unit, 105-icing perception unit, 106-abnormal disturbance perception unit, 107-temporary storage unit, 108-data output interface, 109-wireless transmission unit, 110-database unit, 111-machine learning unit, 112-airborne system update unit, 201-airborne detection system box, 202-airborne data input interface, 203-airborne data processing unit, 204-airborne icing condition perception unit, 205-airborne icing perception unit, 206-airborne abnormal disturbance perception unit, 207-airborne temporary storage unit, 208-airborne data output interface, 209-airborne wireless transmission unit. DETAILED DESCRIPTION

[0020] In order to make the technical solutions and advantages of the embodiments of the present invention more clearly understood, the exemplary embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than an exhaustive list of all the embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0021] Embodiment 1, combination Figure 1 In this embodiment, an ice detection method based on the performance change of a rotating component is described, comprising the following steps: S1. Conduct rotor icing performance change test on rotor test bench in icing wind tunnel, obtain flight parameters and rotor state performance parameters, construct feature matrix containing all the above parameters, and construct corresponding icing state label according to the actual spray opening and closing conditions of wind tunnel (spray opening means icing, spray closing means no icing); S2. dynamically combining the flight parameters and the rotating component parameters by moving average (data with a window width of 5 seconds), normalization (Z-score normalization method) and dimensionless conversion, storing the processed data in a database, and marking the unmarked data as un-iced data by default; S3. Perform physical judgment. When the total temperature of the incoming flow is ≥273.2K, it is directly judged as a non-icing condition without relying on other parameter analysis; S4. Based on the acceleration data obtained in the rotor test bench, use the decision tree model to identify abnormal disturbance phenomena (identify data with sudden changes in vibration acceleration), and filter the interference data through the change characteristics of the flight parameters and the rotating component parameters; S5. Perform first-order differences (10s time difference) on the flight parameters and the rotor state performance parameters, obtain the time series characteristics of the flight parameters and the rotating component parameters, and reconstruct the feature matrix (including all parameters and the obtained time series characteristic parameters). In particular, for a small sample of test data, select the support vector machine (SVM) classification method as the data learning modeling method (based on the amount of data and feature selection method, it can be used for large amounts of data), perform modeling learning on the training data, and use the learned model to judge the test data or actual flight data to determine whether there is icing, the degree of icing and the speed; In particular, for the learning process, the support vector machine obtains the hyperplane parameters by solving the original optimization problem (i.e., the learning process) and b, by calculating the classification decision function To determine whether it is frozen, x is the constructed feature matrix. This also applies to the icing degree classification and icing speed classification.

[0022] S6. Real-time feedback of recognition status through sound and light signals, allowing manual correction of marks, and maintaining the original judgment when no response is given; S7. storing the flight parameters, the rotating component parameters, the processed data and the identification status in a temporary storage medium in full according to a time series; S8. Based on data integrity, all data in the temporary storage medium is divided into level 0 data, level 1 data, level 2 data and level 3 data; S9. Prioritize extracting the level 3 data, randomly balancing sample categories, and updating the machine learning model.

[0023] Specific, combined Figure 2 It can be seen that from the perspective of equipment installation, the present invention completely abandons the traditional mode of additional installation of special detection equipment, and cleverly relies on the existing rotating parts of the aircraft itself and its built-in flight parameter monitoring system. Through innovative algorithms and intelligent sensor integration, it deeply explores the subtle changes in performance parameters of rotating parts during the icing process, such as changes in rotational resistance caused by ice adhesion, dynamic adjustment of vibration frequency and amplitude, fluctuations in power output efficiency, etc., which are used as the key basis for judging whether there is ice and the degree of ice. In this way, not only the impact of the weight of additional equipment on the aircraft load budget is successfully avoided, but also the overall design process of the aircraft is simplified, the R&D cost and maintenance complexity are greatly reduced, and a cost-effective ice detection path is presented for most aircraft, especially small and medium-sized aircraft with limited load.

[0024] At the same time, the present invention focuses on the special needs of the emerging electric vertical take-off and landing aircraft during the critical stage of vertical take-off and landing. The ice detection system constructed by the present invention shows excellent adaptability. When this type of aircraft enters the vertical take-off and landing mode, the airflow field around the fuselage presents high-intensity turbulence and rapidly changing pressure distribution that are very different from conventional flight. Under this extreme working condition, traditional ice detectors often fail due to the limitations of the installation position. The present invention relies on its unique design concept that is decoupled from the flight attitude, and always focuses on the performance evolution of the rotating parts themselves. No matter what flight attitude the aircraft is in or what complex airflow environment it is in, it can accurately capture ice signals, continuously and stably provide flight operators with reliable ice warnings, and effectively ensure the flight safety of electric vertical take-off and landing aircraft during vertical take-off and landing, filling the gap in the existing technology in this application scenario.

[0025] In S1, the flight parameters include: wind speed, static temperature, static pressure, mean effective diameter of supercooled water droplets MVD and liquid water content LWC; the rotating component parameters include: torque, rotation speed, inclination angle and pitch angle.

[0026] In S4, the abnormal disturbance phenomenon includes: turbulence, bird strike and gust of wind.

[0027] In S8, the level 0 data is invalid data or data that is duplicated with existing data, the level 1 data is low-quality data, the level 2 data is ordinary-quality data, and the level 3 data is high-quality data.

[0028] The updating of the machine learning model specifically includes: carrying out classification learning operations, and updating and improving the machine learning model based on the prediction success rate and the reaction delay rate.

[0029] Embodiment 2, combined Figure 3 and Figure 4 This embodiment is described, an ice detection system based on the performance change of a rotating component, the system comprising: a ground full-function system 1, an airborne detection system 2, a ground full-function system box 101 and an airborne detection system box 201; The ground full-function system 1 is located inside the ground full-function system box 101; The airborne detection system 2 is located inside the airborne detection system box 201; The ground-based fully functional system 1 is used to continuously optimize the machine learning model through in-depth analysis and data accumulation; The airborne detection system 2 is used for rapid response, data filtering and implementation of local decision-making; The ground full-function system 1 includes: a data input interface 102, a data processing unit 103, an icing condition sensing unit 104, an icing sensing unit 105, an abnormal disturbance sensing unit 106, a temporary storage unit 107, a data output interface 108, a wireless transmission unit 109, a database unit 110, a machine learning unit 111, and an onboard system update unit 112; The data input interface 102 is used to input the flight parameters and rotating component parameters of the aircraft to be tested; The data processing unit 103 is used to perform dynamic combination processing on the flight parameters and the rotating component parameters by moving average, standardization and dimensionless conversion, store the processed data in a database, and mark the unmarked data as un-iced data by default; The icing condition sensing unit 104 is used to directly determine that the inflow total temperature is ≥ 273.2K as a non-icing condition without relying on other parameter analysis; The abnormal disturbance sensing unit 106 is used to identify abnormal disturbance phenomena using a machine learning model and filter disturbance data according to the change characteristics of the flight parameters and the rotating component parameters; The icing sensing unit 105 is used to determine whether there is icing, the degree of icing and the speed by using a machine learning model based on the time series characteristics of the flight parameters and the rotating component parameters; The data output interface 108 and the wireless transmission unit 109 are used to provide real-time feedback of the recognition status through sound and light signals, allowing manual correction of the mark, and maintaining the original judgment when there is no response; The temporary storage unit 107 is used to store the flight parameters, the rotating component parameters, the processed data and the identification status in a temporary storage medium in full according to a time sequence; The database unit 110 is used to classify all data in the temporary storage medium into level 0 data, level 1 data, level 2 data and level 3 data according to data integrity; The machine learning unit 111 and the airborne system update unit 112 are used to preferentially extract the level 3 data, randomly balance sample categories, and update the machine learning model.

[0030] The airborne detection system 2 includes: an airborne data input interface 202, an airborne data processing unit 203, an airborne icing condition sensing unit 204, an airborne icing sensing unit 205, an airborne abnormal disturbance sensing unit 206, an airborne temporary storage unit 207, an airborne data output interface 208, and an airborne wireless transmission unit 209; The onboard data input interface 202 is used to input the flight parameters and rotating component parameters of the aircraft to be tested; The onboard data processing unit 203 is used to perform dynamic combination processing on the flight parameters and the rotating component parameters by moving average, standardization and dimensionless conversion, store the processed data in a database, and mark the unmarked data as un-iced data by default; The airborne icing condition sensing unit 204 is used to directly determine that the airflow is in a non-icing condition when the total temperature of the incoming airflow is ≥ 273.2K, without relying on other parameter analysis; The onboard abnormal disturbance sensing unit 206 is used to identify abnormal disturbance phenomena using a machine learning model and filter disturbance data according to the change characteristics of the flight parameters and the rotating component parameters; The airborne icing sensing unit 205 is used to determine whether icing, the degree of icing and the speed by using a machine learning model based on the time series characteristics of the flight parameters and the rotating component parameters; The onboard data output interface 208 and the onboard wireless transmission unit 209 are used to provide real-time feedback of the recognition status through sound and light signals, allowing manual correction of the mark and maintaining the original judgment when there is no response; The onboard temporary storage unit 207 is used to store the flight parameters, the rotating component parameters, the processed data and the identification status in a temporary storage medium in full according to a time series.

[0031] Specifically, the airborne detection system 2 exchanges data with the ground full-function system 1 through the wireless transmission unit 209, and the two work together to realize the division of labor mode of "lightweight front end + intelligent back end", balancing the conflicting requirements of real-time, cost and accuracy.

[0032] Figure 5 By comparing the real icing scene with the system detection results, the significant advantages of this method in adaptability to complex environments and detection accuracy are intuitively demonstrated, providing a reliable technical guarantee for the safe operation of aircraft.

Claims

1. An ice detection method based on performance changes of rotating parts, characterized in that: The following steps are involved: S1. Input the flight parameters and rotating component parameters of the aircraft to be tested; S2. dynamically combining the flight parameters and the rotating component parameters by moving average, standardization and dimensionless conversion, storing the processed data in a database, and marking unmarked data as un-iced data by default; S3. When the total temperature of the incoming flow is ≥273.2K, it is directly determined as a non-icing condition without relying on other parameter analysis; S4. Using a machine learning model to identify abnormal disturbance phenomena, and filtering interference data by the change characteristics of the flight parameters and the rotating component parameters; S5. Based on the time series characteristics of the flight parameters and the rotating component parameters, a machine learning model is used to determine whether there is icing, the degree of icing and the speed; S6. Real-time feedback of recognition status through sound and light signals, allowing manual correction of marks, and maintaining the original judgment when no response is given; S7. storing the flight parameters, the rotating component parameters, the processed data and the identification status in a temporary storage medium in full according to a time series; S8. Based on data integrity, all data in the temporary storage medium is divided into level 0 data, level 1 data, level 2 data and level 3 data; S9. Prioritize extracting the level 3 data, randomly balancing sample categories, and updating the machine learning model.

2. The ice detection method based on the performance change of rotating parts according to claim 1 is characterized in that: In S1, the flight parameters include: wind speed, static temperature, static pressure, mean effective diameter of supercooled water droplets MVD and liquid water content LWC output by the wind tunnel measurement and control system; the rotating component parameters include: torque, rotation speed, inclination angle, and pitch angle.

3. The ice detection method based on the performance change of rotating parts according to claim 1, characterized in that: In S4, the abnormal disturbance phenomenon includes: turbulence, bird strike and gust of wind.

4. The ice detection method based on the performance change of rotating parts according to claim 1, characterized in that: In S8, the level 0 data is invalid data or data that is duplicated with existing data, the level 1 data is low-quality data, the level 2 data is ordinary-quality data, and the level 3 data is high-quality data.

5. The ice detection method based on the performance change of rotating parts according to claim 1, characterized in that: The updating of the machine learning model specifically includes: carrying out classification learning operations, and updating and improving the machine learning model based on the prediction success rate and the reaction delay rate.

6. An ice detection system based on performance changes of rotating parts, characterized in that: The system comprises: a ground full-function system (1), an airborne detection system (2), a ground full-function system box (101) and an airborne detection system box (201); The ground full-function system (1) is located inside the ground full-function system box (101); The airborne detection system (2) is located inside the airborne detection system box (201); The ground-based fully functional system (1) is used to continuously optimize the machine learning model through in-depth analysis and data accumulation; The airborne detection system (2) is used for rapid response, data filtering and implementation of local decision-making; The ground full-function system (1) comprises: a data input interface (102), a data processing unit (103), an icing condition sensing unit (104), an icing sensing unit (105), an abnormal disturbance sensing unit (106), a temporary storage unit (107), a data output interface (108), a wireless transmission unit (109), a database unit (110), a machine learning unit (111), and an onboard system update unit (112); The data input interface (102) is used to input flight parameters and rotating component parameters of the aircraft to be tested; The data processing unit (103) is used to perform dynamic combination processing on the flight parameters and the rotating component parameters by moving average, standardization and dimensionless conversion, store the processed data in a database, and mark unmarked data as un-iced data by default; The icing condition sensing unit (104) is used to directly determine that the condition is a non-icing condition when the total temperature of the incoming flow is ≥273.2K, without relying on other parameter analysis; The abnormal disturbance sensing unit (106) is used to identify abnormal disturbance phenomena using a machine learning model, and filter disturbance data according to the change characteristics of the flight parameters and the rotating component parameters; The icing sensing unit (105) is used to determine whether icing occurs, the degree of icing, and the speed by using a machine learning model based on the time series characteristics of the flight parameters and the rotating component parameters; The data output interface (108) and the wireless transmission unit (109) are used to provide real-time feedback of the recognition status through sound and light signals, allowing manual correction of the mark, and maintaining the original judgment when there is no response; The temporary storage unit (107) is used to store the flight parameters, the rotating component parameters, the processed data and the identification status in a temporary storage medium in full according to a time sequence; The database unit (110) is used to classify all data in the temporary storage medium into level 0 data, level 1 data, level 2 data and level 3 data according to data integrity; The machine learning unit (111) and the onboard system updating unit (112) are used to preferentially extract the level 3 data, randomly balance sample categories, and update the machine learning model.

7. The ice detection system based on the performance change of rotating parts according to claim 6, characterized in that: The airborne detection system (2) comprises: an airborne data input interface (202), an airborne data processing unit (203), an airborne icing condition sensing unit (204), an airborne icing sensing unit (205), an airborne abnormal disturbance sensing unit (206), an airborne temporary storage unit (207), an airborne data output interface (208), and an airborne wireless transmission unit (209); The onboard data input interface (202) is used to input flight parameters and rotating component parameters of the aircraft to be tested; The onboard data processing unit (203) is used to perform dynamic combination processing on the flight parameters and the rotating component parameters by moving average, standardization and dimensionless conversion, store the processed data in a database, and mark unmarked data as un-iced data by default; The airborne icing condition sensing unit (204) is used to directly determine that the airflow is in a non-icing condition when the total temperature of the incoming airflow is ≥273.2K, without relying on other parameter analysis; The onboard abnormal disturbance sensing unit (206) is used to identify abnormal disturbance phenomena using a machine learning model, and filter disturbance data according to the change characteristics of the flight parameters and the rotating component parameters; The onboard icing sensing unit (205) is used to determine whether icing occurs, the degree of icing, and the speed using a machine learning model based on the time series characteristics of the flight parameters and the rotating component parameters; The onboard data output interface (208) and the onboard wireless transmission unit (209) are used to provide real-time feedback of the recognition status through sound and light signals, allowing manual correction of the mark, and maintaining the original judgment when there is no response; The onboard temporary storage unit (207) is used to store the flight parameters, the rotating component parameters, the processed data and the identification status completely in a temporary storage medium according to a time sequence.

Images