CN-121998404-A - Dynamic interpretable evaluation method, device and equipment for multi-source data fusion

CN121998404ACN 121998404 ACN121998404 ACN 121998404ACN-121998404-A

Abstract

The application discloses a dynamic interpretable evaluation method, a device and equipment for multi-source data fusion, relates to the technical field of civil aviation operation safety factor evaluation, and is used for solving the technical problems of poor effectiveness, accuracy and reliability in the prior art. The method comprises the steps of adopting a multi-source data acquisition module to acquire target multi-source data, adopting a federal feature alignment module to perform feature alignment on the target multi-source data to acquire multi-source alignment data, inputting the multi-source alignment data into a dynamic evaluation model module to acquire a dynamic weight vector, and adopting an interpretability module to perform space-time risk evolution according to the dynamic weight vector to acquire a visual risk evaluation map. Therefore, the application can improve the effectiveness, accuracy, reliability and the like of civil aviation safety evaluation through the techniques of federal feature alignment technology, dynamic evaluation model of a bidirectional gating mechanism and the like.

Inventors

PAN WEIJUN
FENG YUJIANG
WANG RUNDONG
WANG XUAN
LUAN TIAN
ZUO QINGHAI
JIANG YANQIANG
LI YINXUAN

Assignees

中国民用航空飞行学院

Dates

Publication Date: 20260508
Application Date: 20251217

Claims (10)

1. A method of dynamically interpretable evaluation of multisource data fusion, the method comprising: The method comprises the steps of adopting a multi-source data acquisition module to acquire data to obtain target multi-source data, wherein the target multi-source data comprise real-time radar monitoring data, ADS-B track data, weather information, flight plan FPL and airport operation data; Performing feature alignment on the target multi-source data by adopting a federal feature alignment module to obtain multi-source alignment data, wherein a feature alignment algorithm based on a federal countermeasure network is arranged in the federal feature alignment module; Inputting the multisource alignment data into a dynamic evaluation model module to obtain a dynamic weight vector, wherein the dynamic evaluation model in the dynamic evaluation model module is a bidirectional gating circulation unit BIGRU with an attention mechanism; and according to the dynamic weight vector, carrying out space-time risk evolution by adopting an interpretability module to obtain a visual risk assessment map.
2. The method of claim 1, wherein the step of acquiring the target multi-source data using the multi-source data acquisition module comprises: Adopting a multi-source data acquisition module to acquire data to obtain initial multi-source data; and preprocessing the initial multi-source data to obtain the target multi-source data, wherein the preprocessing comprises outlier detection and missing value interpolation.
3. The method of claim 1, wherein the step of employing a federal feature alignment module to feature align the target multi-source data to obtain multi-source alignment data comprises: performing feature alignment on the target multi-source data by adopting a target optimal transfer function in a federal feature alignment module to obtain multi-source alignment data, wherein the target optimal transfer function is expressed by adopting the following formula: Wherein, the In order to minimize the function G, ; To combat losses; And The characteristic distribution of the source domain and the target domain respectively; is the Wasserstein distance between the source domain and the target domain; Is a coefficient for balancing the countering loss and the distribution difference loss.
4. The method of claim 1, wherein the step of inputting the multi-source alignment data into a dynamic assessment model module to obtain a dynamic weight vector comprises: Carrying out space-time feature extraction on the multisource alignment data by adopting BiGRU-attribute layers of a dynamic evaluation model to obtain a plurality of space-time features; Processing the plurality of space-time features by adopting a multi-head attention unit of the preset dynamic evaluation model to obtain an original weight vector; and determining a dynamic weight vector by adopting the original weight vector, the flow density and a preset weight adjustment function.
5. The method of claim 4, wherein the predetermined weight adjustment function is expressed using the following formula: Wherein, the Adjusting a function for a preset weight; Original weights for the ith feature; The dynamic weight of the ith feature, and n is the total number of features.
6. The method of claim 4, wherein the step of obtaining the visual risk assessment map using the interpretive module for space-time risk evolution based on the dynamic weight vector comprises: performing feature contribution degree quantification of game theory on the plurality of space-time features by adopting an interpretability module to obtain contribution values of all the space-time features; Determining a risk value of an airspace region in a target time period according to the dynamic weight vector, the contribution value of each space-time characteristic and a preset risk value function; and carrying out space-time risk evolution on the risk value of the airspace region in the target time period to obtain the visual risk assessment map.
7. The method of claim 1, wherein the predetermined risk value function is expressed using the following formula: Wherein, the A risk value expressed as a position (x, y) at time t; for the dynamic weights of the ith spatio-temporal feature, The contribution of the ith spatiotemporal feature at position (x, y) and time t.
8. A dynamic interpretable evaluation device for multisource data fusion, the device comprising: The multi-source data acquisition unit is used for acquiring data by adopting a multi-source data acquisition module to obtain target multi-source data, wherein the target multi-source data comprises real-time radar monitoring data, ADS-B track data, weather information, flight plan FPL and airport operation data; The federal feature alignment unit is used for carrying out feature alignment on the target multi-source data by adopting a federal feature alignment module to obtain multi-source alignment data, wherein a feature alignment algorithm based on a federal countermeasure network is arranged in the federal feature alignment module; The dynamic evaluation unit is used for inputting the multi-source alignment data into a dynamic evaluation model module to obtain a dynamic weight vector, wherein the dynamic evaluation model in the dynamic evaluation model module is a bidirectional gating circulation unit BIGRU with an attention mechanism; And the interpretability unit is used for carrying out space-time risk evolution by adopting the interpretability module according to the dynamic weight vector to obtain a visual risk assessment map.
9. An electronic device, the device comprising: a memory for storing program instructions; a processor for invoking program instructions stored in the memory and for performing the method of any of claims 1-7 in accordance with the obtained program instructions.
10. A storage medium having stored thereon computer executable instructions for causing a computer to perform the method of any one of claims 1-7.

Description

Dynamic interpretable evaluation method, device and equipment for multi-source data fusion Technical Field The application relates to the technical field of civil aviation operation safety factor evaluation, and provides a method, a device and equipment for dynamically interpretable evaluation of multi-source data fusion. Background As is well known, the precise and reliable civil aviation safety assessment technology has irreplaceable key roles in preventing risks, guaranteeing flight safety and improving overall operation efficiency. It runs through all links of airport operation, flight scheduling and flight process, and is an important basic stone for ensuring the safe travel of hundreds of millions of passengers. However, the existing civil aviation security assessment methods have several disadvantages: (1) The element weight distribution is mainly realized by manually constructing a judgment matrix, and different expert groups have larger subjectivity and inconsistency when evaluating key indexes such as runway invasion risk and the like. And limited by expert knowledge boundaries, it is difficult to capture all influencing factors comprehensively. (2) A large amount of data, such as ACARS messages, QAR flight data, and weather radar information, are generated during the course of civil aviation operation. However, many airports have the problem of data island, and effective integration and sharing among the key multi-source heterogeneous data cannot be realized, so that an evaluation model faces the dilemma of missing input dimensions in the construction process, and the data utilization rate and the comprehensiveness and accuracy of an evaluation result are affected. (3) The existing civil aviation safety evaluation system mostly adopts a fixed period updating mechanism, lacks an environment adaptation mechanism, and is difficult to respond to dynamic changes in civil aviation operation in time, such as flight changing seasons, extreme weather scenes and the like. In this case, as the environmental change and the scene complexity increase, the accuracy of the evaluation model gradually decreases, and the false alarm rate correspondingly increases, so that the effectiveness and reliability of the evaluation are greatly reduced. Therefore, how to improve the effectiveness, accuracy, reliability and other characteristics of civil aviation safety evaluation becomes a problem to be solved at present. Disclosure of Invention The application provides a dynamic interpretable evaluation method, a device and equipment for multi-source data fusion, which are used for solving the technical problems of poor effectiveness, accuracy and reliability in the prior art. In one aspect, a method for dynamic interpretable evaluation of multisource data fusion is provided, the method comprising: The method comprises the steps of adopting a multi-source data acquisition module to acquire data to obtain target multi-source data, wherein the target multi-source data comprise real-time radar monitoring data, ADS-B track data, weather information, flight plan FPL and airport operation data; Performing feature alignment on the target multi-source data by adopting a federal feature alignment module to obtain multi-source alignment data, wherein a feature alignment algorithm based on a federal countermeasure network is arranged in the federal feature alignment module; Inputting the multisource alignment data into a dynamic evaluation model module to obtain a dynamic weight vector, wherein the dynamic evaluation model in the dynamic evaluation model module is a bidirectional gating circulation unit BIGRU with an attention mechanism; and according to the dynamic weight vector, carrying out space-time risk evolution by adopting an interpretability module to obtain a visual risk assessment map. Optionally, the step of acquiring the target multi-source data by adopting the multi-source data acquisition module includes: Adopting a multi-source data acquisition module to acquire data to obtain initial multi-source data; and preprocessing the initial multi-source data to obtain the target multi-source data, wherein the preprocessing comprises outlier detection and missing value interpolation. Optionally, the step of performing feature alignment on the target multi-source data by using a federal feature alignment module to obtain multi-source aligned data includes: performing feature alignment on the target multi-source data by adopting a target optimal transfer function in a federal feature alignment module to obtain multi-source alignment data, wherein the target optimal transfer function is expressed by adopting the following formula: Wherein, the In order to minimize the function G,;To combat losses; And The characteristic distribution of the source domain and the target domain respectively; is the Wasserstein distance between the source domain and the target domain; Is a coefficient for balancing the countering loss and the distri