Flight safety performance evaluation method and device fusing physiological data and flight quality data
By integrating physiological data and flight quality data into an evaluation method and using a temporal convolutional network to build a personalized safety performance evaluation model, we can solve the blind spot of ignoring physiological status in traditional evaluation methods, achieve a more scientific and comprehensive flight safety assessment, and improve the level of aviation safety management.
Patent Information
- Application Number
- CN202510780181.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-12
- Publication Date
- 2025-09-12
AI Technical Summary
Existing aviation safety assessment methods ignore the potential impact of fluctuations in pilots' physiological states, resulting in incomplete and lack of objectivity in the assessment, making it difficult to scientifically assess flight safety risks.
An evaluation method that integrates physiological data and flight quality data is adopted. Through four levels of devices including acquisition, transmission, analysis and display, a time series convolutional network is used to build a safety performance evaluation level prediction model, and personalized evaluation is achieved through model fine-tuning strategies.
It improves the accuracy and authenticity of flight safety assessments, identifies potential risks, provides scientific basis and intelligent decision-making support, promotes the scientific and intelligent nature of flight training, and improves the safety level of the aviation industry.
Smart Images
Figure CN120634260A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of aviation safety performance evaluation and personalized training, and in particular to a flight safety performance evaluation method and device that integrates physiological data and flight quality data. Background Art
[0002] In the aviation field, flight safety has always been the cornerstone for ensuring the smooth progress of flight missions. Traditional assessment methods mainly rely on the monitoring and analysis of flight phase parameters, but often ignore the potential impact of fluctuations in the pilot's physiological state before flight. The current mainstream assessment methods have limitations: first, they are based only on the analysis of objective flight parameters. Although the data is objective, it ignores the influence of physiological factors, forming a blind spot; second, they rely on subjective reports from pilots, which lacks objectivity. Both methods make it difficult to comprehensively assess safety risks. Therefore, there is an urgent need to explore a personalized safety assessment method and corresponding technical means that integrates pilot physiological data (such as pre-flight status) and objective flight parameters (such as flight quality data) to ensure flight safety in a more scientific and comprehensive manner. Summary of the Invention
[0003] To solve the above technical problems, the present invention proposes a flight safety performance evaluation method and device that integrates physiological data and flight quality data. The present invention is implemented using the following technical solutions:
[0004] A flight safety performance evaluation method and device that integrates physiological data and flight quality data consists of four layers: acquisition, transmission, analysis, and display. The acquisition layer includes a physiological data acquisition device and a flight quality data acquisition device. The transmission layer mainly aggregates data through wireless transmission technologies such as Bluetooth technology and satellite communication technology. The analysis layer is supported by an Internet of Things gateway and a cloud computing platform. The display layer is mainly displayed through terminals such as enterprise safety monitoring platforms, computers, and mobile phones.
[0005] The physiological data acquisition device of the acquisition layer includes an equipment monitoring module, an environment monitoring module, and a sensor wireless transmission acquisition module.
[0006] The equipment monitoring module includes a pressure sensor, an accelerometer (ACC), a photoplethysmography sensor (PPG), a vibration sensor, a nine-axis inertial sensor, etc.
[0007] The environmental monitoring module includes an ALS ambient light sensor, a temperature sensor, etc.
[0008] The sensor wireless transmission acquisition module includes various sensor modules, an industrial-grade single-chip microcomputer, a wireless module, an LED light status indication module, a USB serial port communication interface, a power module, a liquid crystal display module, and a wireless transmission development program interface connected by electrical signals; wherein, the various sensor modules are connected to the wireless transmission module, and the industrial-grade single-chip microcomputer is connected to the wireless transmission module, the LED light status indication module, the USB serial port communication interface, the power module, the liquid crystal display module, and the wireless transmission development program interface by electrical signals.
[0009] The flight quality data acquisition device at the acquisition layer mainly includes: airborne data acquisition modules, airborne wireless communication equipment, satellites, and ground-to-air communication base stations. The airborne data acquisition modules include flight parameter sensors, engine parameter sensors, avionics system sensors, vibration sensors, and fuel level sensors.
[0010] The airborne wireless communication equipment includes a wireless transceiver, a communication control system, an interface and connection device, a power supply and power supply system, a safety protection system, a housing and an installation device.
[0011] The wireless transceiver includes an antenna, a radio frequency front end, a signal processing unit and other parts.
[0012] The communication control system includes a communication protocol stack, channel management, power control, etc.
[0013] The emergency system includes an audible and visual alarm module.
[0014] The communication system includes a wireless transmission module, a digital-to-analog conversion module, and a processor module.
[0015] The transmission layer includes a wireless transmission module and a processor module.
[0016] The analysis layer includes an IoT gateway and computing equipment for cloud computing data processing.
[0017] The display layer includes terminals such as security monitoring platforms, computers, and mobile phones.
[0018] First, the pilots' physiological data was collected before flight training. Experts used the results of quantitative analysis of flight quality data from previous flights to assess safety performance. This data was labeled to construct a labeled dataset. Then, based on a Python-designed temporal convolutional network, the collected physiological data was used to construct a baseline model for predicting the safety performance assessment level. This baseline model was fine-tuned using individual physiological data through a model fine-tuning strategy. Finally, the personalized safety performance assessment level prediction model was used to analyze the pilots' physiological state before flight training to predict their flight training performance. The specific steps are as follows:
[0019] 1) Collect physiological data and flight quality data
[0020] Physiological data is collected by pressure sensors, accelerometers (ACC), photoplethysmography (PPG), vibration sensors, and nine-axis inertial sensors. Flight quality data is collected by flight parameter sensors, engine parameter sensors, avionics system sensors, vibration sensors, and fuel level sensors. The physiological data indicator set is shown in Table 1. Each physiological indicator is a comprehensive result calculated by combining relevant basic physiological signals recorded in real time by sensors with a multimodal algorithm model.
[0021] Table 1 Physiological data indicator set
[0022]
[0023]
[0024] 2) Data transmission
[0025] The data transmission is wireless data transmission. The sensor wireless transmission acquisition module transmits the collected signal to the control center through the Internet of Things gateway and Bluetooth gateway.
[0026] 3) Quantitative evaluation based on flight quality data
[0027] Table 2 Flight quality data index set
[0028]
[0029] The flight quality data transmitted from step 2 are quantitatively analyzed as shown in Table 2. Experts assess the pilots' safety performance and label their physiological data before historical flight training.
[0030] 4) Construct a baseline model for predicting safety performance evaluation levels
[0031] Based on step 3), the labeled physiological data was first cleaned and feature extracted using data preprocessing techniques. The data was then divided into a training set and a test set in an 8:2 ratio. A Python-based temporal convolutional network algorithm was designed to train and construct a baseline model for predicting pilot safety performance evaluation levels. The specific steps are as follows:
[0032] (1) Data preprocessing: linear interpolation is used for short-term missing data, and long-term missing data are marked as invalid segments and removed. Box plots can be used to identify and remove outliers. Moving average or Gaussian filtering is used to smooth the data. The data is then normalized, including maximum and minimum normalization:
[0033]
[0034] (2) Data partitioning: the first 80% of the newly constructed data feature set is used as the training set and the last 20% is used as the test set in the order of timestamps.
[0035] (3) Design the network model. TCN contains three core components: causal convolution, dilated convolution, and residual block.
[0036] The causal convolution ensures that the model can only use current and past inputs to predict the current output, thereby maintaining the causal relationship of the time series. The formula is as follows:
[0037]
[0038] Where: y t is the tth element of the output sequence, x t-i is the ti-th element of the input sequence, w i is the weight of the convolution kernel, k is the size of the convolution kernel, and b is the bias term.
[0039] Dilated convolution increases the receptive field by inserting holes between convolution kernel elements, thereby capturing long-term dependencies. Formula:
[0040]
[0041] Where d is the dilation rate.
[0042] The residual block alleviates the gradient vanishing problem through skip connections and improves the training efficiency of the model. The calculation formula is as follows:
[0043] y=Activation(x+F(x))
[0044] Among them, x is the input, F(x) is a function composed of two convolutional layers, and Activation is the activation function (such as ReLU).
[0045] The parameter settings follow a hierarchical design, such as num_channels = [64, 64, 64] (3 layers, 64 channels per layer), convolution kernel: 3×1 or 5×1 is commonly used to balance local features and computational complexity; Dropout rate: 0.2-0.5 to prevent overfitting (lower values are used for high-dimensional data). The loss function uses multi-class cross entropy:
[0046]
[0047] Among them, y i is the true label, is the predicted probability.
[0048] The optimizer selected was Adam (initial learning rate 3e-4) with learning rate decay. The following metrics were used to evaluate model performance:
[0049] Accuracy:
[0050]
[0051] Weighted F1-Score as evaluation value indicator:
[0052]
[0053] Among them, TP is the sample correctly predicted as the positive class, TN is the sample correctly predicted as the negative class, FP is the sample wrongly predicted as the positive class, and FN is the sample wrongly predicted as the negative class.
[0054] 5) Fine-tune the baseline model using individual physiological data
[0055] Based on the safety performance evaluation level prediction baseline model constructed in step 4), a model fine-tuning strategy is adopted to fine-tune the baseline model using the individual pilot's physiological data to achieve personalized training.
[0056] 6) Output personalized prediction results
[0057] The pilot physiological data collected during the flight training on that day is used and input into the personalized safety performance level assessment prediction model constructed in step 5) to obtain the safety performance level prediction result.
[0058] The beneficial effects of adopting the above technical solution are:
[0059] The present invention constructs a pilot safety performance evaluation level prediction model based on a temporal convolutional network to deeply analyze the pilot's physiological data, uses a model fine-tuning strategy to reduce the impact of individual differences, and improves the authenticity and accuracy of the evaluation. It also combines flight quality data to scientifically evaluate flight training from multiple dimensions, providing a scientific basis for personalized training and evaluation of pilots. At the same time, through monitoring and comprehensive analysis of pilots' physiological and flight quality data, the device can discover and warn of potential flight safety risks, provide intelligent decision-making support for flight management departments, significantly enhance the level of flight safety management, and promote the scientific and intelligent flight training, which is of great significance to improving the safety level and competitiveness of the entire aviation industry. BRIEF DESCRIPTION OF THE DRAWINGS
[0060] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0061] Figure 1 This is an evaluation flow chart of a flight safety performance evaluation method and device that integrates physiological data and flight quality data according to the present invention;
[0062] Figure 2 This is a model construction diagram of a flight safety performance evaluation method and device that integrates physiological data and flight quality data according to the present invention;
[0063] Figure 3 This is a working module structure diagram of a flight safety performance evaluation method and device that integrates physiological data and flight quality data according to the present invention;
[0064] Figure 4 A schematic diagram of the wristband sensor portion of a flight safety performance evaluation method and device integrating physiological data and flight quality data according to the present invention; DETAILED DESCRIPTION
[0065] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention, so that the technical means, creative features, objectives and effects achieved by the present invention can be easily understood.
[0066] See also Figures 1 to 4 A flight safety performance evaluation method and device that integrates physiological data and flight quality data consists of four layers: acquisition, transmission, analysis, and display. The acquisition layer includes a physiological data acquisition device and a flight quality data acquisition device. The transmission layer mainly aggregates data through wireless transmission technologies such as Bluetooth technology and satellite communication technology. The analysis layer is supported by an Internet of Things gateway and a cloud computing platform. The display layer is mainly displayed through terminals such as enterprise safety monitoring platforms, computers, and mobile phones.
[0067] First, the pilots' physiological data before flight training is collected. Experts use the results of quantitative analysis of flight quality data from previous flights to assess safety performance. This historical pre-flight training physiological data is labeled to construct a labeled dataset. Then, a temporal convolutional network-based approach is used to construct a baseline model for predicting the safety performance assessment level using the collected physiological data. This baseline model is then fine-tuned using individual physiological data through a model fine-tuning strategy. Finally, the personalized safety performance assessment level prediction model is used to analyze the pilots' physiological state before flight training to predict their flight training performance. The specific steps are as follows:
[0068] 1) Collect physiological data and flight quality data
[0069] The physiological data is collected by pressure sensors, vibration sensors, nine-axis inertial sensors, etc. The flight quality data is collected by flight parameter sensors, engine parameter sensors, avionics system sensors, vibration sensors, fuel level sensors, etc.
[0070] 2) Data transmission
[0071] The data transmission is wireless data transmission, and the sensor wireless transmission acquisition module transmits the collected signal to the control center through the Internet of Things gateway and the Bluetooth gateway.
[0072] 3) Quantitative evaluation based on flight quality data
[0073] Based on the flight quality data transmitted from step 2, a quantitative analysis is performed, and experts assess the pilots' safety performance and label their physiological data before historical flight training.
[0074] 4) Construct a baseline model for predicting safety performance evaluation levels
[0075] According to step 3), the labeled physiological data are first cleaned and feature extracted using data preprocessing technology, and then divided into training and test sets in a ratio of 8:2. Based on the temporal convolutional network algorithm, a baseline model for predicting the pilot safety performance evaluation level is trained and constructed.
[0076] 5) Fine-tune the baseline model using individual physiological data
[0077] Based on the safety performance evaluation level prediction baseline model constructed in step 4), a model fine-tuning strategy is adopted to fine-tune the baseline model using the individual pilot's physiological data to achieve personalized training.
[0078] 6) Output personalized prediction results
[0079] The pilot physiological data collected during the flight training on that day is used and input into the personalized safety performance level assessment prediction model constructed in step 5) to obtain the safety performance level prediction result.
Claims
1. A flight safety assessment method and device integrating physiological data and flight quality data, comprising: an acquisition layer comprising a physiological data acquisition device and a flight quality data acquisition device; a transmission layer primarily utilizing wireless transmission technologies such as Bluetooth and satellite communication technologies for data aggregation; an analysis layer supported by an Internet of Things gateway and a cloud computing platform; and a display layer primarily utilizing an enterprise safety monitoring platform, a computer, or other terminal for display. The method and device are characterized by: The physiological data acquisition device of the acquisition layer includes equipment monitoring module, environment monitoring module, sensor wireless transmission acquisition module The equipment monitoring module includes a pressure sensor, an accelerometer (ACC), a photoplethysmography sensor (PPG), a vibration sensor, a nine-axis inertial sensor, etc. The environmental monitoring module includes ALS ambient light sensor, temperature sensor, etc. The sensor wireless transmission acquisition module includes various sensor modules, industrial-grade single-chip microcomputer, wireless module, LED light status indicator module, USB serial communication interface, power module, LCD module, wireless transmission development program interface connected by electrical signals; wherein, various sensor modules are connected to the wireless transmission module, the industrial-grade single-chip microcomputer is connected to the wireless transmission module, LED light status indicator module, USB serial communication interface, power module, LCD module, wireless transmission development program interface by electrical signals The flight quality data acquisition device of the acquisition layer mainly includes: airborne data acquisition module, airborne wireless communication equipment, satellite, ground-to-air communication base station. The airborne data acquisition module includes flight parameter sensors, engine parameter sensors, avionics system sensors, vibration sensors, fuel level sensors, etc. The airborne wireless communication equipment includes a wireless transceiver, a communication control system, an interface and a connection device, a power supply and a power supply system, a safety protection system, a housing and a mounting device. The wireless transceiver includes an antenna, a radio frequency front end, a signal processing unit, etc. The communication control system includes communication protocol stack, channel management, power control, etc. The emergency system includes an audible and visual alarm module The communication system includes a wireless transmission module, a digital-to-analog conversion module, and a processor module. The transmission layer includes a wireless transmission module and a processor module The analysis layer includes IoT gateways and computing devices for cloud computing data processing The display layer includes terminals such as security monitoring platforms, computers, and mobile phones.
2. The flight safety assessment method and device integrating physiological data and flight quality data according to claim 1, characterized in that: First, the pilots' physiological data before flight training was collected. At the same time, the safety performance evaluation level was assessed using the results of expert quantitative analysis of flight quality data during previous flights. The physiological data was labeled to construct a labeled dataset. Then, based on the temporal convolutional network method, the collected physiological data was used to construct a baseline model for predicting the safety performance evaluation level. The baseline model was then fine-tuned using the individual's physiological data through a model fine-tuning strategy. Finally, the pilots' physiological state before flight training was analyzed through a personalized safety performance evaluation level prediction model to predict the evaluation level of the pilots' flight training performance.
3. The flight safety performance evaluation method and device integrating physiological data and flight quality data according to claim 1, characterized in that: The specific steps are as follows: 1) Collect physiological data and flight quality data The physiological data is collected by pressure sensors, accelerometers (ACC), photoplethysmography sensors (PPG), vibration sensors, nine-axis inertial sensors, etc. The flight quality data is collected by flight parameter sensors, engine parameter sensors, avionics system sensors, vibration sensors, fuel level sensors, etc. 2) Data transmission The data transmission is wireless data transmission. The sensor wireless transmission acquisition module transmits the collected signal to the control center through the Internet of Things gateway and Bluetooth gateway. 3) Quantitative evaluation based on flight quality data The flight quality data transmitted from step 2 is quantitatively analyzed, and then experts evaluate the pilots' safety performance. The physiological data before historical flight training is labeled as high safety, medium safety, and low safety to form a labeled data set. 4) Construct a baseline model for predicting safety performance evaluation levels According to step 3), the labeled physiological data are first cleaned and feature extracted using data preprocessing technology, and then divided into training set and test set in a ratio of 8:
2. Then, a temporal convolutional network algorithm (TCN) is designed based on Python to train and construct a baseline model A for predicting the pilot safety performance evaluation level. 5) Fine-tune the baseline model using individual physiological data Based on the safety performance evaluation level prediction baseline model A constructed in step 4), a model fine-tuning strategy is adopted to fine-tune the baseline model A using the individual pilot's physiological data. That is, the parameters of the convolutional layer and the temporal feature extraction layer of the TCN network are frozen, and only the last fully connected output layer is adaptively adjusted to achieve personalized training. 6) Output personalized prediction results The physiological data of the pilots collected during the flight training on that day are used and input into the personalized safety performance level evaluation prediction model constructed in step 5) to obtain the safety performance level prediction results.