Why do the authors recommend using a hierarchical method?

While the clustering can be done with a variety of algorithms, the authors recommend using a hierarchical method because of its simplicity and interpretability.

What are the three building blocks in the framework?

Four building blocks interact in the proposed framework: machine comparison, fleet clustering, anomaly detection, and visualization (fig. 1).

Why is the DTW method used in a fleet-based approach?

Due to the fact that the differences appearing in current amplitudes cannot be used in a fleet-based approach (fig. 8), min/max normalization is performed.

What is the cophenetic correlation of a cluster?

A high cophenetic correlation indicates that two or more cluster partitions are present in the data, while a low cophenetic correlation suggests that only a single cluster partition is present in the data.

What are the use-cases of the cophenetic correlation procedure?

All use-casesuse the default implementation of section 3.4, being a (normalized) signal domain representation, machine dissimilarity, clustering, and anomaly scoring.

How do you determine the number of clusters?

In contrast, approaches like hierarchical clustering and DB-scan derive the number of clusters from the structure of the data, often by using a parameter that describes how tight a cluster should be.

What are some examples of applications where all of these assumptions hold?

Practical examples where all of these assumptions hold are applications such as monitoring a conveyor belt or railway components.

Open AccessJournal Article10.1016/J.YMSSP.2019.106585

A general anomaly detection framework for fleet-based condition monitoring of machines

Kilian Hendrickx, +8 more

- 01 May 2020

- Mechanical Systems and Signal Processing

- Vol. 139, pp 106585

TL;DR: This work proposes an unsupervised, generic, anomaly detection framework for fleet-based condition monitoring that uses generic building blocks and offers three key advantages: it allows incorporating domain expertise by user-defined comparison measures, easy interpretability allows a domain expert to validate the predictions made by the framework.

About: This article is published in Mechanical Systems and Signal Processing. The article was published on 01 May 2020. and is currently open access. The article focuses on the topics: Condition monitoring & Environmental pollution.

AI Agents for this Paper

Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps

Most frequently asked questions

1. What are the contributions in "A general anomaly detection framework for fleet-based condition monitoring of machines" ?

This work proposes an unsupervised, generic, anomaly detection framework for fleet-based condition monitoring.

2. What could be used to assess the health of a machine fleet?

Clustering operational parameters could provide relevant data samples of machines in comparable operational conditions, which the framework would then use to assess their current health status’.

3. What is the name of the class of techniques used to detect motor faults?

Motor Current Signature Analysis (MCSA) is a popular class of techniques for detecting motor faults that apply signal processing methods to current signals.

4. What is the traditional approach to condition monitoring?

The traditional condition monitoring approach involves handcrafting an indicator of the phenomena of interest, which requires significant expertise in machine signature analysis and machine usage.

Figure 6: An example of the visualizations provided by the framework, which are analyzed in detail in section 5.3. In this example, drivetrains D1 2 and D2 10 are faulty. Signal data (a.) gives the user a detailed view. In this example, min-max normalization has been applied to the waveforms. A global overview of the fleet is given by pairwise comparisons (b.) and a dendrogram (c.). Finally, the cluster partitions (d.), obtained with thrcc = 0.9, and their corresponding anomaly scores (e.) indicate the prediction for a machine’s health condition. In this example, the framework correctly identifies D1 2 and D2 10 as being anomalous.

Figure 16: Visualization showing the three accelerometer frequency domain signals (a.), machine similarities (b.), clustering (c. & d.) and anomaly scores (e.). The red vertical lines in (a.) indicate the used harmonics 3 – 6. Clusters are partitioned with thrcc = 0.9. Each machine is running stationary at 820 RPM, with D1 2 and D2 10 having a voltage unbalance. D1 2 is less affected by the voltage unbalance compared to D2 10, which is visible in the dendrogram.

Table 8: Performance of the proposed framework for fault detection using vibration signature analysis via harmonic amplitudes. For reference, the last column shows the performance for the classic diagnostics approach (table 7) with the highest F1-score. In five out of the six cases, the proposed approach has the highest F1 score.

Figure 11: DTW of a healthy (blue) and faulty (red) machine. While the measured current signals have a clear difference in shape (a), DTW can non-linearly align them so the signals become similar (b).

Table 6: Performance of the proposed framework for fault detection using electrical signature analysis via spectrogram analysis. Scores corresponding to a scenario’s highest F1-score are indicated in bold. For reference, the last column shows the performance for the classic diagnostics approach (table 3) with the highest F1-score. In five out of the six cases, the proposed approach has the highest F1 score.

Citations

•Journal Article•10.1016/J.YMSSP.2021.108105

An explainable artificial intelligence approach for unsupervised fault detection and diagnosis in rotating machinery

Lucas Costa Brito, +3 more

- 15 Jan 2022

- Mechanical Systems and Signal Processing

TL;DR: A new approach for fault detection and diagnosis in rotating machinery is proposed, namely: unsupervised classification and root cause analysis, and a comparison between models used in machine learning explainability: SHAP and Local Depth-based Feature Importance for the Isolation Forest (Local-DIFFI).

...read moreread less

237

Journal Article•10.1016/j.ymssp.2021.108752

Vibration-based anomaly detection using LSTM/SVM approaches

Kilian Vos, +5 more

- 01 Apr 2022

- Mechanical Systems and Signal Processing

TL;DR: In this paper , a deep learning method with LSTM architectures combined with a one-class support vector machine (SVM) is used to separate abnormal data from normal vibration signals collected during an endurance test of a reduction gearbox and helicopter test flight data measured by multiple sensors situated at different locations of the aircraft.

...read moreread less

142

•Journal Article•10.1016/J.JMSY.2020.05.013

Machine learning for anomaly detection and process phase classification to improve safety and maintenance activities

Elena Quatrini, +3 more

- 01 Jul 2020

- Journal of Manufacturing Systems

TL;DR: A real case study in the pharmaceutical industry validates the proposed anomaly detection methodology, using a 10 months database of 16 process parameters from a granulation process, validating its absolute necessity.

...read moreread less

•Journal Article•10.1186/S10033-021-00570-7

Challenges and Opportunities of AI-Enabled Monitoring, Diagnosis & Prognosis: A Review

Zhibin Zhao, +5 more

- 01 Dec 2021

- Chinese Journal of Mechanical Engineerin...

TL;DR: In this article, the authors provide a systematic overview of the current development, common technologies, open source datasets, codes, and challenges of AI-enabled PHM methods from three aspects of monitoring, diagnosis, and prognosis.

...read moreread less

•Journal Article•10.3390/S21175739

A Survey on Data-Driven Predictive Maintenance for the Railway Industry.

Narjes Davari, +5 more

- 26 Aug 2021

- Sensors

TL;DR: In this paper, the authors present a survey of existing ML and DL techniques for handling predictive maintenance in the railway industry and discuss the main approaches for this specific application within a taxonomy defined by the type of task.

...read moreread less

...

Expand

References

•Book

Artificial Intelligence: A Modern Approach

Stuart Russell, +1 more

- 01 Jan 2020

TL;DR: In this article, the authors present a comprehensive introduction to the theory and practice of artificial intelligence for modern applications, including game playing, planning and acting, and reinforcement learning with neural networks.

...read moreread less

21.4K

Journal Article•10.1080/01621459.1963.10500845

Hierarchical Grouping to Optimize an Objective Function

Joe H. Ward

- 01 Mar 1963

- Journal of the American Statistical Asso...

TL;DR: In this paper, a procedure for forming hierarchical groups of mutually exclusive subsets, each of which has members that are maximally similar with respect to specified characteristics, is suggested for use in large-scale (n > 100) studies when a precise optimal solution for a specified number of groups is not practical.

...read moreread less

19.8K

Journal Article•10.1145/1541880.1541882

Anomaly detection: A survey

Varun Chandola, +2 more

- 30 Jul 2009

- ACM Computing Surveys

TL;DR: This survey tries to provide a structured and comprehensive overview of the research on anomaly detection by grouping existing techniques into different categories based on the underlying approach adopted by each technique.

...read moreread less

11.9K

•Journal Article•10.1007/S11222-007-9033-Z

A tutorial on spectral clustering

Ulrike von Luxburg

- 01 Dec 2007

- Statistics and Computing

TL;DR: In this article, the authors present the most common spectral clustering algorithms, and derive those algorithms from scratch by several different approaches, and discuss the advantages and disadvantages of these algorithms.

...read moreread less

11.1K

•Journal Article•10.1016/0004-3702(96)00007-0

Artificial intelligence: A modern approach

Nils J. Nilsson

- 01 Apr 1996

- Artificial Intelligence

10.1K

...

Expand

A general anomaly detection framework for fleet-based condition monitoring of machines

Chat with Paper

AI Agents for this Paper

Most frequently asked questions

1. What are the contributions in "A general anomaly detection framework for fleet-based condition monitoring of machines" ?

2. What could be used to assess the health of a machine fleet?

3. What is the name of the class of techniques used to detect motor faults?

4. What is the traditional approach to condition monitoring?

5. Why do the authors recommend using a hierarchical method?

6. What are the three building blocks in the framework?

7. Why is the DTW method used in a fleet-based approach?

8. What is the cophenetic correlation of a cluster?

9. What are the use-cases of the cophenetic correlation procedure?

10. How do you determine the number of clusters?

11. What are some examples of applications where all of these assumptions hold?

Figures

Citations

An explainable artificial intelligence approach for unsupervised fault detection and diagnosis in rotating machinery

Vibration-based anomaly detection using LSTM/SVM approaches

Machine learning for anomaly detection and process phase classification to improve safety and maintenance activities

Challenges and Opportunities of AI-Enabled Monitoring, Diagnosis & Prognosis: A Review

A Survey on Data-Driven Predictive Maintenance for the Railway Industry.

References

Artificial Intelligence: A Modern Approach

Hierarchical Grouping to Optimize an Objective Function

Anomaly detection: A survey

A tutorial on spectral clustering

Artificial intelligence: A modern approach

Related Papers (5)

Artificial Intelligent Diagnosis and Monitoring in Manufacturing.

Time-Series Classification Methods: Review and Applications to Power Systems Data

Incorporating Expert Knowledge into a Self-Organized Approach for Predicting Compressor Faults in a City Bus Fleet

Automated Identification of Parameters in Control Systems of Machine Tools

Machining sensor data management for operation-level predictive model