Wyniki wyszukiwania - BazTech

1

A Semi-Supervised Siamese Network for Complex Aircraft System Fault Detection with Limited Labeled Fault Samples

Xinyun Zhu, Sun Jianzhong, Hu Hanchun, Li Chunhua

Eksploatacja i Niezawodność

|

2023

|

Vol. 25, no. 4

art. no. 174382

EN

Health monitoring and fault detection of complex aircraft systems are paramount for ensuring reliable and efficient operation. The availability of monitoring data from modern aircraft onboard sensors provides a wealth of big data for developing deep learning-based fault detection methods. However, aircraft onboard systems typically have limited labeled fault samples and large amounts of unlabeled data. To better utilize the information contained in limited labeled fault samples, a deep learning-based semi-supervisedfault detection method is proposed, which leverages a small number of labeled fault samples to enhance its performance. A novel sample pairing strategy is introduced to improve algorithm performance by iteratively utilizing fault samples. A comprehensive loss function is employed to accurately reconstruct normal samples and effectively separate fault samples. The results of a case study using real data from a commercial aircraft fleet demonstrate the superiority of the proposed method over existing techniques, with improvements of approximately 16.7% in AP, 9.5% in AUC, and 19.2% in F1 score. Ablation studies confirm that performance can be further improved by incorporating additional labeled fault samples during training. Furthermore, the algorithm demonstrates good generalization ability.

2

Performance measurement with high-performance computer using HW-GA anomaly-detection algorithms for streaming data

Fondaj Jakup, Hasani Zirije, Krrabaj Samedin

Computer Science

|

2022

|

T. 23 (3)

395--410

EN

Anomaly detection for streaming real-time data is very important; more significant is the performance of an algorithm in order to meet real-time requirements. Anomaly detection is very crucial in every sector because, by knowing what is going wrong with data/digital systems, we can make decisions to help in every sector. Dealing with real-time data requires speed; for this reason, the aim of this paper is to measure the performance of our proposed Holt–Winters genetic algorithm (HW-GA) as compared to other anomaly-detection algorithms with a large amount of data as well as to measure how other factors such as visualization and the performance of the testing environment affect the algorithm’s performance. The experiments will be done in R with different data sets such as the as real COVID-19 and IoT sensor data that we collected from Smart Agriculture Libelium sensors and e-dnevnik as well as three benchmarks from the Numenta data sets. The real data has no known anomalies, but the anomalies are known in the benchmark data; this was done in order to evaluate how the algorithm works in both situations. The novelty of this paper is that the performance will be tested on three different computers (in which one is a high-performance computer); also, a large amount of data will be used for our testing, as will how the visualization phase affects the algorithm’s performance.

3

Discovery of significant intervals in time-series data

Srinivasan A., Chakravarthy S., Sheshta S.

Foundations of Computing and Decision Sciences

|

2006

|

Vol. 31, No. 1

87-102

EN

This paper deals with time-series sensor data to discover significant intervals for prediction. In our project on intelligent home environment, we need to predict agent's actions (tum on/off the appliances) using previously collected sensor data. We propose an approach that replaces the time-series point values with time-series intervals, which represent the characteristics of the data. Time-series data is folded over a periodicity (day, week, etc.) to form intervals and significant intervals are discovered from them that satisfy minimum support and maximum interval-length criteria. By compressing the time-series data and working with intervals, discovery of significant intervals (for prediction) is efficient. Also, sequential mining algorithms perform better if they are modified to work with reduced dataset of significant intervals. In this paper, we present a suite of algorithms for detecting significant intervals, discuss their characteristics, advantages and disadvantages and analyze their performance.