A SVR-based ensemble approach for drifting data streams with recurring patterns

Liu, Jie; Zio, Enrico

doi:10.1016/j.asoc.2016.06.030

Cited by 14 publications

(4 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For processing data streams using support vector regression (SVR), Liu et al [18] introduced an ensemble SVR. This technique generates subsequent models based on the original, where each model corresponds to distinct time segments within the data stream.…”

Section: Online Streaming Data Machine Learningmentioning

confidence: 99%

Dynamic Industrial Optimization: A Framework Integrates Online Machine Learning for Processing Parameters Design

Yao,

Qian

2024

Future Internet

View full text Add to dashboard Cite

We develop the online process parameter design (OPPD) framework for efficiently handling streaming data collected from industrial automation equipment. This framework integrates online machine learning, concept drift detection and Bayesian optimization techniques. Initially, concept drift detection mitigates the impact of anomalous data on model updates. Data without concept drift are used for online model training and updating, enabling accurate predictions for the next processing cycle. Bayesian optimization is then employed for inverse optimization and process parameter design. Within OPPD, we introduce the online accelerated support vector regression (OASVR) algorithm for enhanced computational efficiency and model accuracy. OASVR simplifies support vector regression, boosting both speed and durability. Furthermore, we incorporate a dynamic window mechanism to regulate the training data volume for adapting to real-time demands posed by diverse online scenarios. Concept drift detection uses the EI-kMeans algorithm, and the Bayesian inverse design employs an upper confidence bound approach with an adaptive learning rate. Applied to single-crystal fabrication, the OPPD framework outperforms other models, with an RMSE of 0.12, meeting precision demands in production.

show abstract

Section: Online Streaming Data Machine Learningmentioning

confidence: 99%

Dynamic Industrial Optimization: A Framework Integrates Online Machine Learning for Processing Parameters Design

Yao,

Qian

2024

Future Internet

View full text Add to dashboard Cite

show abstract

“…Ensembles of classifiers have been used to avoid the bias and risk of errors of individual classifiers and improve the diagnostics accuracy and prediction stability, e.g. [42,43]. Ensemble models require post-processing the outcome of individual classifiers so as to generate a consistent prediction, and to this end, again uncertainty quantification is very important.…”

Section: Ensemble Of Classifiersmentioning

confidence: 99%

Improved power transformer condition monitoring under uncertainty through soft computing and probabilistic health index

Aizpurua

Stewart

McArthur

et al. 2019

Applied Soft Computing

View full text Add to dashboard Cite

Condition monitoring of power transformers is crucial for the reliable and cost-effective operation of the power grid. The health index (HI) formulation is a pragmatic approach to combine multiple information sources and generate a consistent health state indicator for asset management planning. Generally, existing transformer HI methods are based on expert knowledge or data-driven models of specific transformer subsystems. However, the effect of uncertainty is not considered when integrating expert knowledge and data-driven models for the system-level HI estimation. With the increased dynamic and non-deterministic engineering problems, the sources of uncertainty are increasing across power and energy applications, e.g. electric vehicles with new dynamic loads or nuclear power plants with de-energized periods, and transformer health assessment under uncertainty is becoming critical for accurate condition monitoring. In this context, this paper presents a novel soft computing driven probabilistic HI framework for transformer health monitoring. The approach encapsulates data analytics and expert knowledge along with different sources of uncertainty and infers a transformer HI value with confidence intervals for decision-making under uncertainty. Using real data from a nuclear power plant, the proposed framework is compared with traditional HI implementations and results confirm the validity of the approach for transformer health assessment.

show abstract

“…Before establishing the method for dynamic reliability assessment and failure prognostics based on monitored data, some general specifications regarding the system considered are presented as follows: [22], the fuel enthalpy failure threshold is calculated by an equation derived from the post-test strain data. The failure threshold of leakage from the first seal of reactor coolant pump in a nuclear power plant is determined by the field experts [13]. Degradation thresholds are calculated in Pope et al by minimizing the total maintenance cost of the monitored component, or to minimize the downtime of the machine due to maintenance on the monitored component.…”

Section: Target System Configurationmentioning

confidence: 99%

System dynamic reliability assessment and failure prognostics

Liu

Zio

2017

Reliability Engineering & System Safety

Self Cite

View full text Add to dashboard Cite

Traditionally, equipment reliability assessment is based on failure data from a population of similar equipment, somewhat giving an average description of the reliability performance of an equipment, not capturing the specificity of the individual equipment. Monitored degradation data of the equipment can be used to specify its behavior, rendering dynamic the reliability assessment and the failure prognostics of the equipment, as shown in some recent literature. In this paper, dynamic reliability assessment and failure prognostics with noisy monitored data are developed for a system composed of dependent components. Parallel Monte Carlo simulation and recursive Bayesian method are integrated in the proposed modelling framework to assess the reliability and to predict the Remaining Useful Life (RUL) of the system. The main contribution of the paper is to propose a framework to estimate the degradation state of a system composed of dependent degradation components whose conditions are monitored (even without knowing the initial system degradation state) and to dynamically assess the system risk and RUL. As case study, a subsystem of the residual heat removal system of a nuclear power plant is considered. The results shows the significance of the proposed method for tailored reliability assessment and failure prognostics.

show abstract

A SVR-based ensemble approach for drifting data streams with recurring patterns

Cited by 14 publications

References 27 publications

Dynamic Industrial Optimization: A Framework Integrates Online Machine Learning for Processing Parameters Design

Dynamic Industrial Optimization: A Framework Integrates Online Machine Learning for Processing Parameters Design

Improved power transformer condition monitoring under uncertainty through soft computing and probabilistic health index

System dynamic reliability assessment and failure prognostics

Contact Info

Product

Resources

About