KR-20260063803-A - SUBSCRIPTION SOFTWARE AND GENERATIVE ARTIFICIAL INTELLIGENCE-BASED AIRCRAFT APPEARANCE AND ENGINE INSPECTION SYSTEM AND METHOD

KR20260063803AKR 20260063803 AKR20260063803 AKR 20260063803AKR-20260063803-A

Abstract

The present invention relates to a subscription-based software and generative artificial intelligence-based aircraft exterior and engine inspection system and method. The subscription-based software and generative artificial intelligence-based aircraft exterior and engine inspection system according to the present invention comprises a memory storing a program for detecting at least one damage for safety inspection of the aircraft exterior and engine using captured optical and thermal images, and a processor for executing the program, wherein the processor analyzes the captured optical and thermal images using a synthetic network of a CNN-based Region Proposal Network, a meta-learning artificial intelligence model, and an explainable artificial intelligence model, and detects and measures damage to the aircraft exterior and damage to the engine.

Inventors

이철희
구태회
김인수

Assignees

주식회사 딥인스펙션

Dates

Publication Date: 20260507
Application Date: 20241031

Claims (13)

A memory storing a program for detecting at least one damage for safety inspection of the aircraft fuselage exterior and engine using captured optical and thermal images; and A processor that executes the above program, The above processor analyzes the captured optical and thermal images using a synthetic network of a CNN-based Region Proposal Network, a meta-learning AI model, and an explainable AI model, and detects and measures damage to the aircraft's exterior and damage to the engine. Subscription-based software and generative AI-based aircraft exterior and engine inspection system.
In paragraph 1, The above processor detects and measures damage to the exterior of the aircraft fuselage, including dents, indentations, scratches, holes, damaged joints, and lightning damage. Subscription-based software and generative AI-based aircraft exterior and engine inspection system.
In paragraph 1, The above processor detects and measures damage to the engine including cracks, end breakage, gouges, burns, dents, depressions, pressure, loss, holes, and edge tears. Subscription-based software and generative AI-based aircraft exterior and engine inspection system.
In paragraph 1, The above processor establishes an interface between two AI models to integrate a CNN-based Region Proposal Network, a meta-learning AI model, and a synthetic network module of an explainable AI model for AI-based detection of damage to the aircraft airframe exterior and engine into one. Subscription-based software and generative AI-based aircraft exterior and engine inspection system.
In paragraph 4, The above processor integrates the CNN-based Region Proposal Network, meta-learning AI model, and explainable AI model into a single synthetic network module, then generates a model through a lightweighting process, and installs the above model on an embedded board equipped with a GPU or NPU mounted on a drone to detect damage to the aircraft's airframe exterior and engine in real time. Subscription-based software and generative AI-based aircraft exterior and engine inspection system.
In paragraph 1, The above processor is a generative artificial intelligence model that artificially generates types of aircraft airframe and engine damage with a low probability of occurrence. Subscription-based software and generative AI-based aircraft exterior and engine inspection system.
In paragraph 6, The above processor artificially creates holes, lightning damage, and loss. Subscription-based software and generative AI-based aircraft exterior and engine inspection system.
In paragraph 1, The above processor creates a summary table of information regarding the type, location, shape, size, and quantity of damage to the aircraft airframe exterior and jet engines by year of safety inspection, creates an external inspection map of the aircraft airframe exterior and engines based on the summary table, calculates the safety grade of the aircraft airframe exterior and engines, and creates an aircraft defect report, while simultaneously automatically generating the opinions of aircraft maintenance technicians through keyword input using AI-based text generation technology. Subscription-based software and generative AI-based aircraft exterior and engine inspection system.
In paragraph 1, The above processor explains in text the reason for the estimated damage to the aircraft fuselage exterior and engine based on a Feature Ablation algorithm, and highlights key features serving as the basis for the estimation of cracks and defects on the image using a heatmap (a contour-shaped heat map). Subscription-based software and generative AI-based aircraft exterior and engine inspection system.
In paragraph 1, The processor detects damage using an RPN-based deep learning algorithm that detects damage to at least one of the aircraft fuselage exterior and engine from the image data, and an SVM-based classifier that improves surface damage detection performance by optimizing the brightness of an image containing surface damage to fit a learned AI model prior to detection, removing point-shaped noise, and then separately classifying only the image containing damage. Subscription-based software and generative AI-based aircraft exterior and engine inspection system.
In paragraph 1, The above processor executes the safety inspection process of the aircraft airframe exterior and engine through a web browser capable of scaling up in response to an increase in the number of concurrent users and tasks. Subscription-based software and generative AI-based aircraft exterior and engine inspection system.
In paragraph 1, The above processor applies a meta-learning technique (Meta FS-DET) that enables damage detection and classification performance even with a small, unbalanced dataset in cases where it is difficult to synthesize aircraft damage images with images of other aircraft, making it difficult to balance the amount of data between classes and difficult to build a large-scale dataset. Subscription-based software and generative AI-based aircraft exterior and engine inspection system.
In paragraph 1, The above processor performs simulations using a multimodal map-based 3D simulator to perform sub-functions such as collision avoidance between the UAV and the aircraft, localization and mapping, obstacle avoidance, and path planning, and the 3D simulator includes functions for setting the locations of the UAV's approach point, obstacles, start, and destination, as well as functions for uploading the entire 2D/3D map, thereby implementing Sim-to-Real generalization. Subscription-based software and generative AI-based aircraft exterior and engine inspection system.

Description

Subscription Software and Generative Artificial Intelligence-Based Aircraft Appearance and Engine Inspection System and Method The present invention relates to a subscription-based software and generative artificial intelligence-based aircraft exterior and engine inspection system and method. According to conventional technology, there is a lack of technology to perform safety inspections of aircraft exteriors and engines using automated processes, such as artificial intelligence algorithms, for rare damages such as aircraft fuselage breakage and punching. When detecting major damage to the aircraft exterior—such as dents, scratches, punched holes, damaged joints, and lightning damage—and other damage—such as lost rivets, damaged rivets, loose screws, poor adhesion, worn paint finish, exterior poor marking, patches, and repairs—based on artificial intelligence algorithms, if rare damage such as fuselage damage or punching is included, the AI model training There is a problem that it is not performed quickly and efficiently. FIG. 1 is a configuration diagram of a SaaS-based SOC facility monitoring/safety diagnosis/inspection dedicated platform according to an embodiment of the present invention. FIG. 2 is a dedicated platform for deep inspection of SaaS-based SOC facilities according to an embodiment of the present invention. FIG. 3 illustrates a CNN-based RPN artificial intelligence model according to an embodiment of the present invention. FIG. 4 is a configuration diagram of an AI solution for detecting and quantifying defects in the superstructure and substructure of a bridge facility based on explainable AI according to an embodiment of the present invention. FIG. 5 is a flowchart of an artificial intelligence model for detecting and quantifying defects in the superstructure and substructure of a bridge facility based on explainable artificial intelligence according to an embodiment of the present invention. FIG. 6 illustrates transfer learning applied to an aircraft fuselage damage detection algorithm according to an embodiment of the present invention. FIG. 7 is an example of an aircraft fuselage damage dataset according to an embodiment of the present invention. FIG. 8 illustrates a flowchart of image synthesis, data augmentation, PCA analysis-based data balancing, pre-training, and fine-tuning according to an embodiment of the present invention. FIG. 9 is an example of an aircraft fuselage damage dataset by damage size according to an embodiment of the present invention. FIG. 10 is an example of an aircraft engine damage dataset according to an embodiment of the present invention. FIG. 11 illustrates the class of aircraft fuselage and engine damage data according to an embodiment of the present invention. FIG. 12 is a diagram of an artificial intelligence model for crack detection/quantification of the superstructure and substructure of a bridge facility according to an embodiment of the present invention. FIG. 13 is a webpage for detecting cracks/defects in the substructure (pier) of a bridge facility according to an embodiment of the present invention. FIG. 14 is a webpage showing the crack detection results of a bridge facility substructure (pier) according to an embodiment of the present invention. FIG. 15 is a webpage showing the result of detecting defects in the substructure (pier) of a bridge facility according to an embodiment of the present invention. FIG. 16 is a webpage showing the crack measurement results of a bridge facility substructure (pier) according to an embodiment of the present invention. Figure 17 is a webpage showing the XAI-Heatmap result of a bridge facility substructure (pier) according to an embodiment of the present invention. FIG. 18 is a webpage showing the XAI-Captioning result of a bridge facility substructure (pier) according to an embodiment of the present invention. FIG. 19 is a configuration diagram of a reinforcement learning model for detecting cracks and defects in the superstructure and substructure of a bridge facility according to an embodiment of the present invention. FIG. 20 is a conceptual diagram of a generative artificial intelligence model according to an embodiment of the present invention. FIG. 21 is a configuration diagram of a generative artificial intelligence model according to an embodiment of the present invention. FIG. 22 is a configuration diagram of a Feature Ablation explainable artificial intelligence model according to an embodiment of the present invention. Figure 23 shows the image output result of a Feature Ablation explainable artificial intelligence model (Center: correct answer, Right: output result). FIG. 24 illustrates the feature ablation visualization process according to an embodiment of the present invention. FIG. 25 is a block diagram showing a computer system for implementing a method according to an embodiment of the present invention. FIG. 26 illustrates a configuration diagram of a meta-learning model (Meta-FSDet) acco