Fault Detection and Diagnosis in the DAMADICS Benchmark Actuator System – A Hidden Markov Model Approach

Almeida, Gilberto Machado de; Park, Song Won

doi:10.3182/20080706-5-kr-1001.02102

Cited by 11 publications

(6 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Of these, deeplearning based techniques have received significant attention for MVTS anomaly detection owing to (a) their ability to scale to high dimensions and model complex patterns in various domains, compared to straightforward statistical approaches such as out of limits approaches [5], [29], [30], (b) fast inference and applicability to streaming time-series typical in CPS unlike many distance-based and pattern-based techniques that are not applicable to streaming time-series, as they require both training and test data during inference [24]- [27], and (c) the ability to localize anomalous time points within sequences, unlike techniques [24], [25], [31] that work at the coarser level to detect anomalous sub-sequences. Anomaly diagnosis has been approached primarily from a supervised classification perspective [32]. In the unsupervised context, four studies [9]- [12] mention that ranking of scores or errors can be used to diagnose the cause of anomalies but only [10] shows experimental results on an open dataset.…”

Section: Related Workmentioning

confidence: 99%

An Evaluation of Anomaly Detection and Diagnosis in Multivariate Time Series

Garg

Zhang

Samaran

et al. 2022

IEEE Trans. Neural Netw. Learning Syst.

View full text Add to dashboard Cite

Several techniques for multivariate time series anomaly detection have been proposed recently, but a systematic comparison on a common set of datasets and metrics is lacking. This paper presents a systematic and comprehensive evaluation of unsupervised and semi-supervised deep-learning based methods for anomaly detection and diagnosis on multivariate time series data from cyberphysical systems. Unlike previous works, we vary the model and post-processing of model errors, i.e. the scoring functions independently of each other, through a grid of 10 models and 4 scoring functions, comparing these variants to state of the art methods. In time-series anomaly detection, detecting anomalous events is more important than detecting individual anomalous time-points. Through experiments, we find that the existing evaluation metrics either do not take events into account, or cannot distinguish between a good detector and trivial detectors, such as a random or an all-positive detector. We propose a new metric to overcome these drawbacks, namely, the composite F-score (F c1), for evaluating time-series anomaly detection.Our study highlights that dynamic scoring functions work much better than static ones for multivariate time series anomaly detection, and the choice of scoring functions often matters more than the choice of the underlying model. We also find that a simple, channel-wise model -the Univariate Fully-Connected Auto-Encoder, with the dynamic Gaussian scoring function emerges as a winning candidate for both anomaly detection and diagnosis, beating state of the art algorithms.

show abstract

Section: Related Workmentioning

confidence: 99%

An Evaluation of Anomaly Detection and Diagnosis in Multivariate Time Series

Garg

Zhang

Samaran

et al. 2022

IEEE Trans. Neural Netw. Learning Syst.

View full text Add to dashboard Cite

show abstract

“…Huang (2008) suggested a similar (see Section 2.1) HMM approach to sensor problem diagnosis but limited considerations to faults in the output channels and input signals taking values from a finite, discrete set. Almeida and Park (2008) learned an HMM corresponding to each operating condition and, unlike the approach proposed in this work, does not make use of the process model. There, fault detection is achieved by a classification scheme that chooses the HMM that maximizes the probability of a given sequence of observations.…”

Section: Fault Modeling Using Hidden Markov Modelsmentioning

confidence: 99%

Fault Detection in Process Systems using Hidden Markov Disturbance Models

Wong

Lee

2009

IFAC Proceedings Volumes

View full text Add to dashboard Cite

“…Since then, others fields have experienced success, namely bioinformatics [6], telecommunications [7] and financial engineering [8], to mention a few. Studies applying HMM in chemical process monitoring are given by [9][10][11][12][13], to mention a few. The case study is the DAMADICS actuator benchmark problem, which is commonly used for developing and comparison of monitoring systems [14].…”

Section: Introductionmentioning

confidence: 99%

“…The detection of the latter mode, more usual in practice, is harder. An initial application of HMM in this benchmark was previously carried out by the same authors [9,15]. The present work introduces an improved methodology, geared to periodic processes, that leads to better results.…”

Section: Introductionmentioning

confidence: 99%

Fault Detection in Continuous and Periodic Industrial Chemical Processes with Hidden Markov Models

Almeida¹,

Park²

2016

Modelling, Simulation and Identification / 841: Intelligent Systems and Control

View full text Add to dashboard Cite

The development of automatic and reliable fault detection systems is still a challenge nowadays. Chemical processes are complex by nature by presenting non linear dynamics, multiple modes with constant interchanges, and spatial and serial correlations, to mention a few. To address these issues, this work explores the hidden Markov model (HMM) technique to construct a fault detection system for continuous and periodic processes. The DAMADICS actuator benchmark, with thirty four abrupt fault scenarios, was used for evaluation purposes. Abrupt faults of low magnitude are challenging and of great interesting in practice. The results obtained with the proposed methodology were compared to classical multivariate statistical process control (MSPC) techniques. They show a significant higher performance leading to earlier fault detection given a fixed false alarm rate of 1%.

show abstract

Fault Detection and Diagnosis in the DAMADICS Benchmark Actuator System – A Hidden Markov Model Approach

Cited by 11 publications

References 18 publications

An Evaluation of Anomaly Detection and Diagnosis in Multivariate Time Series

An Evaluation of Anomaly Detection and Diagnosis in Multivariate Time Series

Fault Detection in Process Systems using Hidden Markov Disturbance Models

Fault Detection in Continuous and Periodic Industrial Chemical Processes with Hidden Markov Models

Contact Info

Product

Resources

About