Predictive maintenance solution for industrial systems - an unsupervised approach based on log periodic power laws

Tracking #: 860-1840

Authors:

NameORCID
Bogdan LobodzinskiORCID logo https://orcid.org/0000-0001-9452-6078


Responsible editor: 

Tobias Kuhn

Submission Type: 

Research Paper

Abstract: 

A new unsupervised predictive maintenance analysis method based on the renormalization group approach used to discover critical behavior in complex systems has been proposed. The algorithm analyzes univariate time series and detects critical points based on a newly proposed theorem that identifies critical points using a Log Periodic Power Law function fits. Application of a new algorithm for predictive maintenance analysis of industrial data collected from reciprocating compressor systems is presented. Based on the knowledge of the dynamics of the analyzed compressor system, the proposed algorithm predicts valve and piston rod seal failures well in advance.

Manuscript: 

Tags: 

  • Reviewed

Data repository URLs: 

Date of Submission: 

Monday, July 22, 2024

Date of Decision: 

Wednesday, January 15, 2025


Nanopublication URLs:

Decision: 

Undecided

Solicited Reviews:


2 Comments

meta-review by editor

The reviewers agree that this manuscript is promising but they also point out to a number of valid issues still to be resovled, in particular with respect to background / related work. We therefore ask for a revision of the manuscript taking these points into account.

Tobias Kuhn (https://orcid.org/0000-0002-1267-0234)

Review document content of Review #1

The review document of Review #1 is not accessible due to a bug in the system. Therefore I paste its content here:

Summary of paper in a few sentences:
The study presents an unsupervised predictive maintenance analysis that leverages data collected from reciprocating compressor systems and time series data to identify critical behaviors through Renormalization Group (RG) analysis and models their effects on the time series using the Log Periodic Power Law (LPPL) function to predict failures in advance. However, it has been understood that further effort is required in some points. The points open for improvement in the article are summarized below.
Reasons to revisions:
1. In the Introduction section, the problem presented in the study should be referenced more clearly. (page number 1, line numbers 28-44, and page number, line numbers 1-4). For example, 'This would suggest that unsupervised approaches may prove to be a better-suited tool for building PM processes' – according to whom?
2. The mathematical foundation and adaptation of the analysis framework to the data have been explained sufficiently. They have shown that the proposed method performs well.
3. What are the characteristics of the data collected through this process? Why are unsupervised approaches suitable for predictive maintenance? Although this is somewhat explained in the Related Works section, it is not sufficiently clear. It should be elaborated further with more examples linked to previous studies.
4. A citation should be provided (p.2, lines 24-28).
5. 'It is sufficient to determine whether it can be determined whether a given point in the time series is the initial (initiating) moment of IB or not.' A citation should be provided.
6. However, there is insufficient information on other studies conducting predictive maintenance analysis and the methods used. The need for such a framework has not been adequately discussed based on previous works. For instance, what does the proposed method achieve that the Statistical Process Control method does not? A methodological framework has been presented without sufficiently considering the advantages and disadvantages of existing methods.