Ethiopian Airlines' Boeing 737-8 MAX nosedived and crashed shortly after takeoff on March 10, 2019, at Ejere Town, south of Addis Ababa. A faulty angle of attack (AOA) sensor was the cause of the crash. Many airpl...
详细信息
ISBN:
(纸本)9781665478960
Ethiopian Airlines' Boeing 737-8 MAX nosedived and crashed shortly after takeoff on March 10, 2019, at Ejere Town, south of Addis Ababa. A faulty angle of attack (AOA) sensor was the cause of the crash. Many airplane accidents have been linked to faulty AOA sensors in the past. The majority of the AOA sensor fault detection, isolation, and accommodation (SFDIA) literature relied on linear model-driven techniques, which are not suitable when the system's model is uncertain, complex, or nonlinear. Traditional multilayer perceptron (MLP) models have been employed in data-driven models in the literature and the effectiveness of deep learning-based data-driven models has not been investigated. In this work, a data collection and processing method that ensures the collected data is not monotonous and a data-driven model for AOA SFDIA is proposed. The proposed model uses a deep learning-based recurrent neural network (RNN) to accommodate for faulty AOA measurement under flight conditions with faulty AOA measurement, faulty total velocity measurement, and faulty pitch rate measurement. Conventional residual analysis with a fixed threshold is used to detect and isolate faulty AOA sensors. The proposed and benchmark models are trained with the adaptive momentum estimation (Adam) algorithm. We show that the proposed model effectively detects, isolates, and accommodates faulty AOA measurements when compared to other data-driven benchmark models. The method is able to detect and isolate faulty AOA sensors with a detection delay of 0.5 seconds for ramp failure and 0.1 seconds for step failure.
Learning effective representations of source code is critical for any Machine Learning for Software Engineering (ML4SE) system. Inspired by natural language processing, large language models (LLMs) like Codex and Code...
详细信息
ISBN:
(纸本)9798350329964
Learning effective representations of source code is critical for any Machine Learning for Software Engineering (ML4SE) system. Inspired by natural language processing, large language models (LLMs) like Codex and CodeGen treat code as generic sequences of text and are trained on huge corpora of code data, achieving state of the art performance on several software engineering (SE) tasks. However, valid source code, unlike natural language, follows a strict structure and pattern governed by the underlying grammar of the programming language. Current LLMs do not exploit this property of the source code as they treat code like a sequence of tokens and overlook key structural and semantic properties of code that can be extracted from code-views like the control Flow Graph (CFG), data Flow Graph (DFG), Abstract Syntax Tree (AST), etc. Unfortunately, the process of generating and integrating code-views for every programming language is cumbersome and time consuming. To overcome this barrier, we propose our tool COMEX - a framework that allows researchers and developers to create and combine multiple code-views which can be used by machine learning (ML) models for various SE tasks. Some salient features of our tool are: (i) it works directly on source code (which need not be compilable), (ii) it currently supports Java and C#, (iii) it can analyze both method-level snippets and program-level snippets by using both intra-procedural and inter-procedural analysis, and (iv) it is easily extendable to other languages as it is built on tree-sitter - a widely used incremental parser that supports over 40 languages. We believe this easy-to-use code-view generation and customization tool will give impetus to research in source code representation learning methods and ML4SE. The source code and demonstration of our tool can be found at https://***/IBM/tree- sitter-codeviews and https://***/GER6U87FVbU, respectively.
High-precision digital twin model construction and real-time data fusion technology are the core of digital twin technology, which jointly support the whole life cycle simulation, verification, prediction and control ...
详细信息
ISBN:
(数字)9798331536169
ISBN:
(纸本)9798331536176
High-precision digital twin model construction and real-time data fusion technology are the core of digital twin technology, which jointly support the whole life cycle simulation, verification, prediction and control of physical entities. In this paper, the construction process of high-precision digital twin model is deeply discussed, including four stages: data acquisition and processing, preliminary model establishment, model refinement and optimization, simulation verification and iterative update. In terms of key technologies and algorithms, this paper analyzes 3D modeling and geometric processing technology, physical properties and behavior simulation technology, and data fusion and real-time update technology. In particular, this paper discusses the role of data fusion algorithms such as Kalman filter in improving the accuracy and reliability of data. In the part of real-time data fusion technology, this paper discusses the data source and diversity, the fusion method and process, and the strategy of using clustering algorithm and machine learning model to improve the efficiency of data fusion. In the face of challenges, this paper points out some problems, such as data security and privacy protection, insufficient generalization ability of models, difficulties in data acquisition and integration, bottlenecks in real-time performance and lack of industry standards, and puts forward corresponding improvement measures. Finally, this paper looks forward to the future development of digital twin technology, and emphasizes the importance of data-driven intelligent modeling methods, cross-industry standardization and innovative applications in intelligent manufacturing, smart cities, smart medical care and other fields.
We illustrate challenges and preliminary results of the FoAIming project (artificial intelligence and robotics in polymer foaming) that aims at exploiting robotics and artificial intelligence (AI) methods in the field...
详细信息
This paper proposes an intelligent dynamic modeling method for strip rolling process. Actuators of a cold rolling mill perform actions, including work roll bending, intermediate roll bending, and roll gap tilting, to ...
This paper proposes an intelligent dynamic modeling method for strip rolling process. Actuators of a cold rolling mill perform actions, including work roll bending, intermediate roll bending, and roll gap tilting, to regulate the product quality. Conventional first-principles modeling necessitates intricate mathematical equations derived from processanalysis and action evaluation. However, the complexity of the model increases the difficulty of upgrading the model. To address this limitation, a novel approach involving a deep neural network with a Gaussian distribution layer is proposed to effectively capture the process dynamics. A negative log likelihood is established as the loss function for the model using real-world industrial data. Experiment outcomes demonstrate the efficacy of the proposed method in achieving an accurate dynamic model.
The article presents the development and experimental verification of a control system based on a highlevel controller by a group of unmanned aerial vehicles (UAVs) based on the ROS 2 framework and the PX4 autopilot i...
详细信息
ISBN:
(数字)9798331512194
ISBN:
(纸本)9798331512200
The article presents the development and experimental verification of a control system based on a highlevel controller by a group of unmanned aerial vehicles (UAVs) based on the ROS 2 framework and the PX4 autopilot in the Gazebo simulation environment. The system is designed to address challenges related to autonomous navigation, group coordination, and obstacle avoidance in complex environments. The modular architecture of the control system is described, which includes such components as a high-level controller, a Gcommand processing module, a flight control module, and a data visualization module. For modeling and testing the system, the Gazebo simulation environment is used Gazebo, which provides realistic physical modeling and integration with ROS 2. The results of modeling and experimental studies of the movement of a group of UAVs are presented, during which the correctness of command execution, coordination of the group of UAVs and the stability of the system to external influences are checked.
In computational pathology, whole slide images represent the primary data source for AI-driven diagnostic algorithms. However, due to their high resolution and large size, these images undergo a patching phase. In thi...
详细信息
The missile fuze control system, is the most central functional component of the strategic missile control system. In the trend of increasingly strong combat target protection, how to improve the accuracy of detonatio...
详细信息
Providing reliable information on human activities and behaviors is an extremely important goal in various application areas such as healthcare, entertainment, and security. Within the working environment, a correct i...
详细信息
ISBN:
(数字)9783031060182
ISBN:
(纸本)9783031060182;9783031060175
Providing reliable information on human activities and behaviors is an extremely important goal in various application areas such as healthcare, entertainment, and security. Within the working environment, a correct identification of the actual performed tasks can provide an effective support in the assessment of the risk associated to the execution of the task itself, and thus preventing the development of work-related musculoskeletal diseases. In this perspective, wearable-based Human Activity Recognition systems have been representing a prominent application. This study aimed to compare three different classification approaches appointed from supervised learning techniques, namely k-Nearest Neighbors, Support Vector Machine and Decision Tree. Motion data, related to several working activities realized in the large-scale retail distribution, were collected by using a full-body system based on 17 Inertial Measurement Units(MVN Analyze, XSens). Reliable features in both time- and frequency-domain were first extracted from raw 3D accelerations and angular rates data, and further processed by Principal Component analysis, with 95% threshold. The classification models were validated via 10-fold cross-validation on a defined training dataset. k-Nearest Neighbors classifier, which provide the best results on the training session, was eventually tested for generalization on additional data acquired on few specific tasks. As a result, considering 5 main macro activities, k-Nearest Neighbors provided a classification accuracy of 80.1% and a computational time of 1865.5 s. To test the whole assessment process, the activities labelled by the classification model as handling of low loads at high frequency were automatically evaluated for risk exposure via OCRA Checklist method.
With the continuous growth in the number of Android applications and the size of their codebases, it has become increasingly difficult for testers to manually analyze and trigger the functionalities of interest in eac...
详细信息
暂无评论