Pilot health index estimation method and system based on plateau flight process

Abstract

The invention discloses a pilot health index pre-estimation method and system based on a plateau flight process, and relates to the technical field of pilot health index pre-estimation, and the method comprises the steps: obtaining the current flight data in the flight time, inputting the current flight data into various types of pre-estimation models for processing, obtaining each piece of health pre-estimation information corresponding to the duration of the flight time in the future; according to each piece of health pre-estimation information, the model weight of each type of pre-estimation model is calculated by using an objective weighting method, and a health index pre-estimation result is calculated by using each type of model weight and the corresponding health pre-estimation index; obtaining new current flight data of the duration corresponding to each time step according to each time step interval, and updating the future duration and the health index estimation result through the new current flight data; and real-time monitoring and estimation of the health indexes of the whole plateau flight process of the pilot are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of pilot health index prediction, and more particularly, to a pilot health index prediction method and system based on a plateau flight process. Background Art

[0002] The harsh plateau environment poses great risks to civil aviation operations. Pilots are the core element in ensuring the safety of civil aviation operations, and the health and safety of plateau civil aviation pilots are crucial to plateau civil aviation operations. Current health and safety protection measures for plateau pilots are mainly based on "pre-job" health checks. Existing measures cannot achieve real-time detection of the health status of pilots during plateau flights, let alone pre-perception of health risks, posing a potential danger to the safety of plateau civil aviation operations. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for estimating pilot health indicators based on the plateau flight process, so as to realize real-time monitoring and estimation of pilot health indicators throughout the entire plateau flight process, and to realize real-time early warning and disposal of health risks through the evolution estimation of health indicators.

[0004] The above technical objectives of the present invention are achieved through the following technical solutions: In a first aspect, the present application provides a method for estimating pilot health indicators based on a high-altitude flight process, comprising the following specific steps: Acquire flight status data from multiple pilots during the flight process and use the flight status data to build a training dataset. The flight status data includes environmental indicator data for each flight phase, as well as the health indicator data of the pilots corresponding to the environmental indicator data in each flight phase. The corresponding flight phases, environmental indicator data, and health indicator data in the training data set are used as a single training set, and the training data set is used to form multiple types of training samples with different time steps and different time step intervals. The training samples include multiple training sets; For each type of training sample, the training sample is input into the corresponding preset initial model and trained until the preset initial model reaches the preset training end condition, and the preset initial model that reaches the training end condition is determined as the estimated model; When the flight time of the flight process reaches the minimum time step in each type of training sample, the current flight data within the flight time is obtained. The current flight data includes flight phase information, environmental index information, and health index information. The current flight data is input into each type of prediction model for processing to obtain various health prediction information corresponding to the duration of the flight time in the future; Based on the health prediction information of each type, the objective weighting method is used to calculate the model weights of each type of prediction model. The health indicator prediction results are calculated using the model weights of each type and the corresponding health prediction indicators. New current flight data corresponding to the duration of each time step is obtained according to each time step interval. The future duration and health indicator prediction results are updated using the new current flight data.

[0005] The beneficial effects of the present invention are as follows: in this solution, first, a training data set is established by acquiring the flight status data of multiple pilots throughout the entire flight process, wherein the established training data set includes environmental index data and health index data of each flight stage during the flight process, and the environmental index data is also the real physical index of the pilot; and the environmental index data and health index data at the same time in each flight stage are used to form a single training set, and for the training data set, by setting different time steps and different time step intervals, multiple types of training samples are obtained, and each training sample includes multiple training sets, wherein the flight process can be regarded as a time period, and the above-mentioned time steps are intercepted from the time period, preferably, the time steps are arranged from short to long, and the longest time step can be the same as the entire flight time period; the above-mentioned time step interval represents the time interval between each training set in each time step; secondly, the corresponding preset initial model is trained respectively using each type of training sample, that is, the type of training sample that exists The total number is the same as the number of preset initial models. For the training process, the input of the preset initial model is the health indicators, environmental data, and flight phase of the time interval within each time step, and the output is the health indicator data for the future time interval. The time step interval is delayed for a certain period of time. After the delay, the time interval of the time step is input for training again until the training of each preset initial model is completed. Secondly, when the actual flight process reaches the minimum time step, the health indicators, environmental data, and flight phase within the time step are obtained and input into each estimation model for processing to obtain the health estimation information for the future time step. Since the time intervals of each model are different during training, the health estimation information output by each estimation model is weighted to obtain the final health indicator estimation result for a certain period of time in the future. Finally, according to the setting of the time step interval, the health indicators, environmental data, and flight phase of the historical time length are continuously input according to the interval time, and the health indicator estimation results for the future time length and the future time length are updated through the same weighted calculation.

[0006] In this proposal, a pilot health index estimation method is provided. By constructing an estimation model for pilots' health indicators throughout the entire plateau flight process, a method for improving model estimation accuracy by combining parallel estimation and objective weighting is proposed to achieve accurate evolution estimation of health indicators at different stages throughout the plateau flight process.

[0007] On the basis of the above technical solution, the present invention can also be improved as follows.

[0008] Furthermore, the above method further includes: Performing a health risk level assessment based on the values of each indicator in the health indicator estimation result, obtaining a risk level assessment result corresponding to the health indicator estimation result, and processing the pilot using a preset method corresponding to the risk level assessment result; The beneficial effects of adopting the above further scheme are: by constructing a health risk warning method for plateau civil aviation pilots in different flight stages, and proposing targeted decision-making and disposal methods based on different risk levels.

[0009] Furthermore, the above method further includes: Preprocess the training data set. Preprocessing includes outlier processing and data normalization. Outlier processing is specifically as follows: ; Where, Indicates outliers The processed value, n Indicates the abnormal indicator The total amount of data corresponding to the flight phase, Indicates the first i data values; Normalized processing is: ; Where, represents the data after normalization. represents the data before normalization. Indicates the minimum and maximum values of the indicator during the flight phase to which the indicator corresponds.

[0010] Furthermore, the above preset initial model is specifically: ; Where, express t The estimated health indicators of pilots predicted at all times, Represents the weight matrix between the intermediate layer and the output layer of the preset initial model, Indicates the activation function of the preset initial model, select tanh As the activation function, Represents the weight matrix between the intermediate layers of the preset initial model, express t-1 The pilot's health indicator data at all times, Represents the input data matrix, which is a single training set formed by corresponding flight phases, environmental indicator data, and health indicator data. Represents the bias term between the intermediate layers of the preset initial model, Represents the bias term between the intermediate layer and the output layer of the preset initial model, Represents the weight matrix between the input layer and the intermediate layer in the preset initial model.

[0011] Furthermore, the above training termination condition is that the loss function of the preset initial model does not exceed the threshold for multiple times. The loss function is specifically: ; Where, L represents the value of the loss function, y Represents the pilot's health indicator data during the flight, f(x) Represents the estimated value of the preset initial model output, Indicates the control parameters of the preset initial model.

[0012] Furthermore, the above model weights are specifically: ; Where, Indicates the first j The model weights of the estimated models, Indicates the first j The standard deviation of the health prediction index obtained by the prediction model is Indicates the first j The degree of conflict between an estimation model and other estimation models, n Indicates the total number of estimated model types.

[0013] Furthermore, the estimated results of the above health indicators are as follows: ; Where, Represents the estimated results of health indicators, Indicates the n The model weights of the estimated models of each type, Indicates the n Health estimation indicators for each type of estimation model, n Indicates the total number of estimated model types.

[0014] In a second aspect, the present application provides a pilot health index prediction system based on a plateau flight process, which is applied to a pilot health index prediction method based on a plateau flight process according to any one of the first aspects, including: The first module is used to obtain flight status data of multiple pilots during the flight process and establish a training data set based on the flight status data. The flight status data includes environmental index data for each flight phase and the health index data of the pilots corresponding to the environmental index data in each flight phase; The second module is used to use the corresponding flight phases, environmental indicator data, and health indicator data in the training data set as a single training set, and to form multiple types of training samples of the training data set with different time steps and different time step intervals. The training samples include multiple training sets; The third module is used to input the training samples of each type into the corresponding preset initial model and train the model until the preset initial model reaches a preset training end condition, and determine the preset initial model that reaches the training end condition as the estimated model; The fourth module is used to obtain the current flight data within the elapsed time when the elapsed time of the flight process reaches the minimum time step among the training samples of each type. The current flight data includes flight phase information, environmental indicator information, and health indicator information. The current flight data is input into the various types of prediction models for processing to obtain the future health prediction information corresponding to the elapsed time. The fifth module is used to calculate the model weights of various types of estimation models using the objective weighting method based on the health estimation information, and to calculate the health indicator estimation results using the model weights of various types and the corresponding health estimation indicators, and to obtain new current flight data corresponding to the duration of each time step according to each time step interval, and to update the future duration and health indicator estimation results through the new current flight data.

[0015] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any one of the methods in the first aspect when executing the computer program.

[0016] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions enable a computer to execute any one of the methods in the first aspect.

[0017] Compared with the prior art, the present invention has at least the following beneficial effects: In the present application, a training data set is first established by obtaining flight status data of multiple pilots throughout the flight process. The established training data set includes environmental index data and health index data for each flight stage during the flight process. The environmental index data is also the real physical index of the pilot; and the environmental index data and health index data at the same time in each flight stage are used to form a single training set. For the training data set, different time steps and different time step intervals are set to obtain multiple types of training samples, and each training sample includes multiple training sets, wherein the flight process can be regarded as a time period, and the above-mentioned time step is intercepted from the time period. Preferably, the time steps are arranged from short to long, and the longest time step can be the same as the entire flight time period; the above-mentioned time step interval represents the time interval between each training set in each time step; secondly, the corresponding preset initial model is trained respectively using the training samples of each type, that is, the total number of types of existing training samples is the same as the preset initial model. The number of initial models is the same. For the training process, the input of the preset initial model is the health indicators, environmental data, and flight phase of the time interval within each time step, and the output is the health indicator data for the future time interval. The time step interval is delayed for a certain period of time. After the delay, the time interval of the time step is input for training again until the training of each preset initial model is completed. Secondly, when the actual flight process reaches the minimum time step, the health indicators, environmental data, and flight phase within the time step are obtained and input into each estimation model for processing to obtain the health estimation information for the future time step. Since the time intervals of each model are different during training, the health estimation information output by each estimation model is weighted to obtain the final health indicator estimation result for a certain period of time in the future. Finally, according to the setting of the time step interval, the health indicators, environmental data, and flight phase of the historical time length are continuously input according to the interval time, and the health indicator estimation results for the future time length and the future time length are updated through the same weighted calculation.

[0018] In this application, the pilot health index estimation method provided constructs an estimation model for the pilot's health indicators throughout the entire plateau flight process, and proposes a model estimation accuracy improvement method that combines parallel estimation and objective empowerment to achieve accurate evolution estimation of health indicators at different stages throughout the plateau flight process; at the same time, by constructing a health risk warning method for plateau civil aviation pilots in different flight stages, and proposing targeted decision-making and disposal methods based on different risk levels. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings: Figure 1 A flowchart of an estimation method according to an embodiment of the present invention; Figure 2 A connection diagram of an estimation system according to an embodiment of the present invention; Figure 3 Schematic diagram of the connection of electronic equipment in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0022] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0023] In the description of the embodiments of the present invention, "a plurality of" means at least two.

[0024] Example 1: In order to achieve real-time monitoring and prediction of pilot health indicators during the entire plateau flight process, this embodiment provides a pilot health indicator prediction method based on the plateau flight process, such as Figure 1 As shown, the following specific steps are included: S1. Acquire flight status data of multiple pilots during the flight process and establish a training data set based on the flight status data. The flight status data includes environmental indicator data for each flight phase and health indicator data of the pilots corresponding to the environmental indicator data in each flight phase.

[0025] Among them, the flight stages during the flight process can include 7 stages: pre-post, ground pre-flight, push-out taxiing, take-off climb, cruising, descent, and arrival, which can be represented by 0-6 in the training data set; environmental indicator data can include aircraft altitude, aircraft speed, cabin pressure, cabin oxygen content, cabin temperature, etc., which can be collected by the airport meteorological data center before boarding the plane and by the aircraft avionics system after boarding the plane; the pilot's health indicator data can include heart rate, blood oxygen, blood pressure, and body temperature. These data can be obtained through medical-grade smart wearable devices, and the pilot's electrocardiogram can be obtained through medical-grade smart electrocardiogram monitoring stickers.

[0026] Specifically, based on the collected pilot health indicator data, environmental data, and flight phase data, the data is subjected to three processes: data analysis, data grouping, and data preprocessing. In terms of data analysis, the distributed stream processing framework Flink can be used to consume the data stream in Kafka and perform real-time analysis. In data grouping, the parsed data is grouped separately according to the pilot's ID information to obtain individual time series data for each pilot. In terms of data preprocessing, the time series data of each pilot is subjected to data preprocessing work to refine the health and environmental data, including steps such as outlier processing and data normalization.

[0027] Optionally, the above method further includes: Preprocess the training data set. Preprocessing includes outlier processing and data normalization. Outlier processing is specifically as follows: ; Where, Indicates outliers The processed value, n Indicates the abnormal indicator The total amount of data corresponding to the flight phase, Indicates the first i data values.

[0028] Furthermore, the above normalization process is: ; Where, represents the data after normalization. represents the data before normalization. Indicates the minimum and maximum values of the indicator during the flight phase to which the indicator corresponds.

[0029] S2, takes the corresponding flight phases, environmental indicator data and health indicator data in the training data set as a single training set, and forms multiple types of training samples with different time steps and different time step intervals. The training samples include multiple training sets.

[0030] Among them, the training dataset of the real-time prediction model for the whole process of pilots' health indicators during high-altitude flight is the whole-process data of S high-altitude pilots after preprocessing, including "before duty - before ground flight - push-out taxiing - takeoff climb - cruise - descent - arrival". There are 11 parameters in total, including 5 pilot health indicator data (heart rate, blood pressure, blood oxygen, body temperature, and electrocardiogram), 5 flight environment indicator data (aircraft altitude, aircraft speed, cabin pressure, oxygen content in the aircraft cabin, and cabin temperature), and 1 flight state data (divided into 7 states: before duty, before ground flight, push-out taxiing, takeoff climb, cruise, descent, and arrival); then the format of a single training set of the preset initial model can be expressed as (S, 600, 11), where S represents S pilots, 600 represents the number of time-series data collected for each pilot (it can be collected once every 1 minute for 10 hours, a total of 600 data), and 11 represents 11 indicator data; during the training process, 70% of the training dataset can be used as the training dataset, and the data format can be expressed as (0.7S, 600, 11); 30% can be used as the test set, which can be expressed as (0.3S, 600, 11).

[0031] S3. For each type of training sample, input the training sample into the corresponding preset initial model and train it until the preset initial model reaches the preset training end condition, and determine the preset initial model that reaches the training end condition as the prediction model.

[0032] Among them, for the construction of the prediction model: Import the processed whole-process data of the pilot "before duty - before ground flight - push-out taxiing - takeoff climb - cruise - descent - arrival" into the preset initial model for training (the models used here can adopt deep learning algorithms such as RNN recurrent neural network and variational autoencoder VAE), and make the following settings for the preset initial model: 1. The preset initial model is divided into three parts: input, output, and intermediate layer. The number of input and output layers is 1, and the number of intermediate layers ranges from 20 to 40, which can be dynamically adjusted according to the prediction accuracy of the real-time prediction model training process of the pilot. 2. The duration of the input data of each preset initial model can be data of M time steps (60 < M < 120), that is, the health, environment, and flight stage data collected by the pilot in the past M / 60 hours at the current flight moment. 3. Each preset initial model re-inserts the health, environment, and flight stage data collected by the pilot in the past M / 60 hours into the model every N minutes (5 < N < 20), and repeats the training of the preset initial model.

[0033] Optionally, the above preset initial model is specifically: ; Where, express t The estimated health indicators of pilots predicted at all times, Represents the weight matrix between the intermediate layer and the output layer of the preset initial model, Indicates the activation function of the preset initial model, select tanh As the activation function, Represents the weight matrix between the intermediate layers of the preset initial model, express t-1 The pilot's health indicator data at all times, Represents the input data matrix, which is a single training set formed by corresponding flight phases, environmental indicator data, and health indicator data. Represents the bias term between the intermediate layers of the preset initial model, Represents the bias term between the intermediate layer and the output layer of the preset initial model, Represents the weight matrix between the input layer and the intermediate layer in the preset initial model.

[0034] Specifically, the training process of the preset initial model can be optimized using the following methods (but not limited to) (existing conventional model optimization methods): a. Use early stopping to monitor model performance on the validation set to avoid overfitting; b. Add dropout layers between intermediate layers of the preset initial model to reduce overfitting; c. Optimize and adjust model parameters through methods such as random search or Bayesian optimization.

[0035] The process of building the prediction model for pilot health indicators is as follows: ① The health indicators, environmental data, and flight phases (data format is (0.7S, M, 11)) of S pilots in the pre-service phase in the flight database for M time steps (60≤M≤120) are used as input data for the preset initial model; ② The five health index data of S pilots in the flight database at the time of M / 60 hours (data format is (0.7S, M, 5)) are used as the output data of the preset initial model; ③ Every N minutes (5≤N≤20), the health indicators, environmental data, and flight phases of the S pilots preceding the current moment M / 60 hours are used as model inputs, and the health indicator data after the current moment M / 60 hours are used as model outputs to train the preset initial model; ④ When the Huber loss value of each preset initial model reaches the set threshold standard (the Huber value is <0.1 for 10 consecutive times), the training process of the preset initial model is ended, and the model is the final estimation model.

[0036] Furthermore, to improve the accuracy and robustness of the pilot health indicator estimation model, the training process can be divided into 16 batches and conducted in parallel. The difference between the models in different batches lies in the difference in M and N (M: the health indicators, environmental data, and flight phases collected at different times of the pilot; N: the input time interval of the training data). The M and N values of the 16 batches can be set to (M60N5, M60N10, M60N15, M60N20, M80N5, M80N10, M80N15, M80N20, M100N5, M100N10, M100N15, M100N20, M120N5, M120N10, M120N15, M120N20). Finally, the training results are the estimation models corresponding to the 16 preset initial models.

[0037] Furthermore, after the preset initial model training is completed, it can be tested: based on the health indicators, environmental data and flight phase data of S pilots in the flight database (the input data format is (0.3S, 600, 11), and the output data format is (0.3S, 600, 5)), 16 parallel preset initial models are tested. If the Huber loss value of the test process reaches the set threshold standard (the Huber value is <0.1 for 10 consecutive times), the model passes the model test. If the Huber loss value does not reach the set threshold standard, the model fails the model test and needs to be further optimized using the model optimization method.

[0038] Optionally, the above training termination condition is: the loss function of the preset initial model does not exceed the threshold value multiple times, and the loss function is specifically: ; Where, L represents the value of the loss function, y Represents the pilot's health indicator data during the flight, f(x) represents the estimated value of the preset initial model output, Indicates the control parameters of the preset initial model.

[0039] S4. When the flight time of the flight process reaches the minimum time step in each type of training sample, the current flight data within the flight time is obtained. The current flight data includes flight phase information, environmental indicator information, and health indicator information. The current flight data is input into each type of estimation model for processing to obtain various health estimation information corresponding to the duration of the flight time in the future.

[0040] The steps for actually using the above estimation model are as follows: 1. Real-time and continuous data collection: The health and environmental indicators of pilots whose health indicators are to be estimated are collected in real time and continuously for two hours before boarding the aircraft. The data collection indicators and collection methods are as described above, and the collection frequency can be 1 meter / time; 2. Pre-flight health assessment: Based on the pilot's pre-boarding health data, combined with the Civil Aviation Administration of China and airlines' health standards for pilots operating in high altitude areas, the flight will be canceled if the health indicators do not meet the standards. If the indicators meet the standards, a full-process health indicator evolution assessment will be conducted. 3. Health indicator evolution prediction: After the pilot's health and environmental data collection period reaches M / 60 hours (as described above for the M and N settings), the collected data is preprocessed and input into 16 parallel models for estimating the pilot's health indicators throughout the entire high-altitude flight process. This data is used to generate real-time predictions of the pilot's five health indicators over the next M / 60 hours. Thereafter, the input data is updated every N minutes, with the past M / 60 hours of collected data fed into the prediction model to generate real-time predictions of the pilot's health indicator evolution over the next M / 60 hours. 4. Estimation Result Processing: By adjusting the values of M and N, the real-time health indicator prediction model for pilots throughout high-altitude flight, trained with different M and N values, was used to perform real-time health indicator predictions for pilots in 16 batches. Furthermore, based on the real-time health indicator prediction results from these 16 parallel prediction models, the CRITIC objective weighting method was used to assign weights to the 16 models, and the final prediction results were obtained through weighted averaging.

[0041] S5. Based on the health prediction information of each type, the objective weighting method is used to calculate the model weight of each type of prediction model, and the health indicator prediction result is calculated using the model weight of each type and the corresponding health prediction index. According to each time step interval, the new current flight data corresponding to the duration of each time step is obtained, and the future duration and health indicator prediction results are updated through the new current flight data.

[0042] The above model weights are specifically: ; Where, Indicates the first j The model weights of the estimated models, Indicates the first j The standard deviation of the health prediction index obtained by the prediction model is Indicates the first j The degree of conflict between an estimation model and other estimation models, n Indicates the total number of estimated model types.

[0043] Optionally, the above health indicator estimation results are as follows: ; Where, Represents the estimated results of health indicators, Indicates the n The model weights of the estimated models of each type, Indicates the n Health estimation indicators for each type of estimation model, n Indicates the total number of estimated model types.

[0044] Specifically, according to the 16 estimation models, then: , .

[0045] Optionally, the above method further includes: S6, performing a health risk level assessment based on each indicator value in the health indicator estimation result, obtaining a risk level assessment result corresponding to the health indicator estimation result, and processing the pilot using a preset method corresponding to the risk level assessment result.

[0046] Among them, a real-time monitoring platform for pilot health status can be built, and the real-time collected plateau pilot health index data, environmental data, and pilot health index prediction data can be transmitted to the monitoring platform through the aircraft ATG, so as to realize real-time and continuous monitoring and early warning of pilot health index data; at the same time, when the pilot completes the plateau route flight, the health indicators, environmental data and flight stages collected throughout the process will be re-input into the health index prediction model for training to continuously improve the robustness and accuracy of the model.

[0047] Specifically, for the different stages of plateau flight for plateau pilots, namely "pre-job-ground pre-flight-rollout-takeoff and climb-cruise-descent-arrival", professional aviation doctors can issue health indicator risk evaluation standards and divide the health risks in different flight stages into four risk levels: low, medium, high and extremely high.

[0048] The system combines the real-time prediction results of pilot health indicators with the corresponding health risk assessment level standards for different flight phases, and evaluates the pilot's potential health risk level within 30 minutes. All process and result data are transmitted in real time to the real-time pilot health status monitoring platform, and the pilot health indicator data is displayed in real time: 1. If the assessment result is low risk, continuous monitoring and assessment of health status will be carried out; 2. If the assessment result is medium-to-high risk, the flight doctor will intervene and the frequency of health indicator collection will be increased to 30 seconds. The data will be transmitted to the airline monitoring center in real time via ATG, and the flight doctor will monitor the pilot's health status in real time. 3. If the evaluation result is extremely high risk, the current pilot will be suspended from flying and replaced by a second crew. The frequency of health indicator collection will be increased to 10 seconds per time, and the data will be transmitted in real time to the airline monitoring center via ATG, and the aviation doctor will monitor the pilot's health status in real time.

[0049] Example 2: This embodiment of the present application provides a pilot health index prediction system based on a plateau flight process, which is applied to a pilot health index prediction method based on a plateau flight process in any one of Example 1, such as Figure 2 As shown, including: The first module is used to obtain flight status data of multiple pilots during the flight process and establish a training data set based on the flight status data. The flight status data includes environmental indicator data for each flight phase and the health indicator data of the pilots corresponding to the environmental indicator data in each flight phase.

[0050] The second module is used to use the corresponding flight phases, environmental indicator data and health indicator data in the training data set as a single training set, and to form multiple types of training samples with different time steps and different time step intervals in the training data set. The training samples include multiple training sets.

[0051] The third module is used to input the training samples of each type into the corresponding preset initial model and train until the preset initial model reaches the preset training end condition, and determine the preset initial model that reaches the training end condition as the estimated model.

[0052] The fourth module is used to obtain the current flight data within the flight time when the flight time of the flight process reaches the minimum time step in each type of training sample. The current flight data includes flight phase information, environmental indicator information and health indicator information. The current flight data is input into each type of estimation model for processing to obtain various health estimation information corresponding to the duration of the flight time in the future.

[0053] The fifth module is used to calculate the model weights of various types of estimation models using the objective weighting method based on the health estimation information, and to calculate the health indicator estimation results using the model weights of various types and the corresponding health estimation indicators, and to obtain new current flight data corresponding to the duration of each time step according to each time step interval, and to update the future duration and health indicator estimation results through the new current flight data.

[0054] Example 3: This embodiment of the present application provides an electronic device, such as Figure 3 As shown, it includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any one of the methods in Example 1 is implemented.

[0055] Example 4: An embodiment of the present application provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions enable a computer to execute any one of the methods in Example 1.

[0056] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for estimating pilot health indicators based on plateau flight process, characterized by: The specific steps include: Acquiring flight status data of multiple pilots during a flight and establishing a training dataset using the flight status data, wherein the flight status data includes environmental indicator data for each flight phase and health indicator data of the pilots corresponding to the environmental indicator data for each flight phase; The corresponding flight phases, environmental indicator data, and health indicator data in the training data set are used as a single training set, and the training data set is used to form multiple types of training samples with different time steps and different time step intervals, wherein the training samples include multiple training sets; For each type of training sample, input the training sample into the corresponding preset initial model and train it until the preset initial model reaches a preset training end condition, and determine the preset initial model that reaches the training end condition as the estimated model; When the elapsed flight time of a flight process reaches the minimum time step among the training samples of each type, the current flight data within the elapsed flight time is obtained. The current flight data includes flight phase information, environmental index information, and health index information. The current flight data is input into each type of prediction model for processing to obtain various health prediction information corresponding to the elapsed flight time in the future. Based on the health prediction information of each type, the model weight of each type of prediction model is calculated using the objective weighting method, and the health indicator prediction result is calculated using the model weight of each type and the corresponding health prediction indicator. According to each time step interval, new current flight data corresponding to the duration of each time step is obtained, and the future duration and health indicator prediction results are updated using the new current flight data.

2. The method for estimating pilot health indicators based on plateau flight process according to claim 1 is characterized in that: The method further comprises: A health risk level assessment is performed according to each indicator value in the health indicator estimation result to obtain a risk level assessment result corresponding to the health indicator estimation result, and the pilot is processed using a preset method corresponding to the risk level assessment result.

3. The method for estimating pilot health indicators based on plateau flight process according to claim 1 is characterized in that: The method further comprises: The training data set is preprocessed, and the preprocessing includes outlier processing and data normalization processing, wherein the outlier processing is specifically: ； Where, Indicates outliers The processed value, n Indicates the abnormal indicator The total amount of data corresponding to the flight phase, Indicates the first i data values; The normalization process is: ； Where, represents the data after normalization. represents the data before normalization. Indicates the minimum and maximum values of the indicator during the flight phase to which the indicator corresponds.

4. The method for estimating pilot health indicators based on plateau flight process according to claim 1 is characterized in that: The preset initial model is specifically: ； Where, express t The estimated health indicators of pilots predicted at all times, Represents the weight matrix between the intermediate layer and the output layer of the preset initial model, Indicates the activation function of the preset initial model, select tanh As the activation function, Represents the weight matrix between the intermediate layers of the preset initial model, express t-1 The pilot's health indicator data at all times, Represents the input data matrix, which is a single training set formed by corresponding flight phases, environmental indicator data, and health indicator data. Represents the bias term between the intermediate layers of the preset initial model, Represents the bias term between the intermediate layer and the output layer of the preset initial model, Represents the weight matrix between the input layer and the intermediate layer in the preset initial model.

5. The method for estimating pilot health indicators based on plateau flight process according to claim 1 is characterized in that: The training end condition is: the loss function of the preset initial model does not exceed the threshold value multiple times. The loss function is specifically: ； Where, L represents the value of the loss function, y Represents the pilot's health indicator data during the flight, f(x) represents the estimated value of the preset initial model output, Indicates the control parameters of the preset initial model.

6. The method for estimating pilot health indicators based on plateau flight process according to claim 1, characterized in that: The model weights are specifically: ； Where, Indicates the first j The model weights of the estimated models, Indicates the first j The standard deviation of the health prediction index obtained by the prediction model is Indicates the first j The degree of conflict between an estimation model and other estimation models, n Indicates the total number of estimated model types.

7. The method for estimating pilot health indicators based on plateau flight process according to claim 6 is characterized in that: The health indicator estimation results are as follows: ； Where, Represents the estimated results of health indicators, Indicates the n The model weights of the estimated models of each type, Indicates the n Health estimation indicators for each type of estimation model, n Indicates the total number of estimated model types.

8. A pilot health index prediction system based on plateau flight process, applied to a pilot health index prediction method based on plateau flight process according to any one of claims 1 to 7, characterized in that: include: The first module is configured to obtain flight status data of multiple pilots during a flight and establish a training dataset using the flight status data, wherein the flight status data includes environmental indicator data for each flight phase and health indicator data of the pilots corresponding to the environmental indicator data in each flight phase; The second module is used to use the corresponding flight phase, environmental indicator data and health indicator data in the training data set as a single training set, and to form multiple types of training samples of the training data set with different time steps and different time step intervals, wherein the training samples include multiple training sets; A third module is configured to input the training samples of each type into a corresponding preset initial model and train the model until the preset initial model meets a preset training end condition, and determine the preset initial model that meets the training end condition as the estimated model; The fourth module is used to obtain current flight data within the elapsed flight time when the elapsed flight time reaches the minimum time step among the training samples of each type. The current flight data includes flight phase information, environmental index information, and health index information. The current flight data is input into each type of prediction model for processing to obtain future health prediction information corresponding to the elapsed flight time. The fifth module is used to calculate the model weights of various types of estimation models using the objective weighting method based on the health estimation information, and to calculate the health indicator estimation results using the model weights of various types and the corresponding health estimation indicators, and to obtain new current flight data corresponding to the duration of each time step according to each time step interval, and to update the future duration and health indicator estimation results through the new current flight data.

9. An electronic device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the method according to any one of claims 1 to 7 is implemented when the processor executes the computer program.

10. A non-transitory computer-readable storage medium, characterized in that The non-transitory computer-readable storage medium stores computer instructions, which enable a computer to execute the method of any one of claims 1 to 7.

Images