A5065258230- Software System Performance and Reliability
- Network Security and Intrusion Detection
- Anomaly Detection Techniques and Applications
- Software Reliability and Analysis Research
- Cloud Computing and Resource Management
- Refrigeration and Air Conditioning Technologies
- Heat Transfer and Optimization
- Carbon Dioxide Capture Technologies
- Software Engineering Research
- Advanced Thermodynamics and Statistical Mechanics
- Mechanical Failure Analysis and Simulation
- Data Stream Mining Techniques
- Spacecraft and Cryogenic Technologies
- Advanced Thermodynamic Systems and Engines
- Advanced Battery Technologies Research
- Superconducting Materials and Applications
- Engineering Applied Research
- Engineering Diagnostics and Reliability
- Power Systems and Technologies
- Advanced Algorithms and Applications
- IoT and Edge/Fog Computing
- Advanced Sensor and Control Systems
- Software Testing and Debugging Techniques
- Thermodynamic and Exergetic Analyses of Power and Cooling Systems
Technical Institute of Physics and Chemistry
2014-2024
Chinese Academy of Sciences
2014-2024
Wuhan University
2018-2021
Logs that record system abnormal states (anomaly logs) can be regarded as outliers, and the k-Nearest Neighbor (kNN) algorithm has relatively high accuracy in outlier detection methods. Therefore, we use kNN to detect anomalies log data. However, there are some problems when using anomalies, three of which are: excessive vector dimension leads inefficient algorithm, unlabeled data cannot support imbalance number distorts classification decision algorithm. In order solve these problems,...
Logs play an important role in the maintenance of large-scale systems. The number logs which indicate normal (normal logs) differs greatly from that anomalies (abnormal logs), and two types have certain differences. To automatically obtain faults by K-Nearest Neighbor (KNN) algorithm, outlier detection method with high accuracy, is effective way to detect logs. However, characteristics large scale very uneven samples, will affect results KNN algorithm on log-based anomaly detection. Thus, we...
Exception handling is widely used in software development to guarantee code robustness and system reliability. Developers are expected choose appropriate strategies ensure exceptions handled properly without causing program crashes or unintended behaviors. However, making such choices challenging especially for the novices due lack of experience on exceptional flow design. To assist developers deciding how handle exceptions, we propose a method automatically recommend exception based...
Using the k -nearest neighbor (kNN) algorithm in supervised learning method to detect anomalies can get more accurate results. However, when using kNN anomaly, it is inefficient at finding neighbors from large-scale log data; same time, data are imbalanced quantity, so a challenge select proper for different distributions. In this paper, we propose log-based anomaly detection with efficient selection of and automatic neighbors. First, search based on minhash MVP-tree. The used group similar...
The log analysis-based system fault diagnosis method can help engineers analyze the events generated by system. K-means algorithm perform analysis well and does not require a lot of prior knowledge, but K-means-based needs to be improved in both efficiency accuracy. To solve this problem, we propose based on reclustering algorithm. First, vectorization PV-DM language model obtain low-dimensional vectors which provide effective data support for subsequent diagnosis; then, improve make effect...
Distributed systems have been widely used in the information technology industry. However, with increasing scale and complexity of distributed systems, efficiency accuracy manual anomaly detection system logs decreased. Therefore, there is a great demand for highly accurate efficient automatic method based on log analysis to ensure reliability stability large-scale systems. In this paper, we propose VeLog, an variational autoencoders (VAEs). offline training phase, VeLog learns patterns...
A control system comprised of the structure, key logic, alarm and interlock is presented in this article for using it helium liquefier. The process instrumentation diagram architecture are analysed. With architecture, logic including open-loop control, closed-loop subsection investigated. Finally liquefier intoduced. proposed paper has been successfully applied to developed by TIPC.
Log analysis can be used for software system anomaly detection, and ensemble learning handle log data with imbalanced characteristics. Therefore, log-based detection is a good choice. However, the existing balancing methods in may destroy distribution of original affect accuracy results. Besides, rules do not take into account relationship between samples to detected historical data. we propose method NW (Neighbor Weighting) rules, which uses based on spectral clustering so that balanced...
A diagnostic method for screw compressor faults using vibration signals was introduced and applied to a new type of helium oil-injected in the 80L/h liquefier. The study revealed that destruction oil film due impurity particles could lead rotor rubbing poor meshing, compromising reliability these compressors. Based on frequency-domain analysis, major occurred at meshing frequency rotors series its harmonics. characteristic bearing parts also indicated had slight wear early stage. waveform's...
When SaaS software suffers from the problem of response time degradation, scaling deployment resources that support operation can improve time, but also means an increase in costs due to additional resources. For purpose saving while improving out is alternative approach. However, how affects case important issue for effectively time. Therefore, this paper, we propose a method analysing impact on Specifically, define scaling-out and then leverage queueing theory analyse According conclusions...