Combining Chronicle Mining and Semantics for Predictive Maintenance in Manufacturing Processes

Tracking #: 2173-3386

Qiushi Cao
Ahmed Samet
Cecilia Zanni-Merk
François de Bertrand de Beuvron
Christoph Reich

Responsible editor: 
Guest Editors SemWeb of Things for Industry 4.0 - 2019

Submission type: 
Full Paper
Within manufacturing processes, faults and failures may cause severe economic loss. With the vision of Industry 4.0, artificial intelligence techniques such as data mining play a crucial role in automatic fault and failure prediction. However, data mining especially pattern mining results normally lack both machine and human understandable representation and interpretation of knowledge, bringing obstacles to novice users to interpret the prediction results. To tackle this issue, in this paper we introduce a novel hybrid approach to facilitate predictive maintenance tasks in manufacturing processes. The proposed approach is a combination of data mining and semantics, within which chronicle mining is used to predict the future failures of the monitored industrial machinery, and a Manufacturing Predictive Maintenance Ontology (MPMO) with its rule-based extension is used to predict temporal constraints of failures and to represent the predictive results formally. As a result, Semantic Web Rule Language (SWRL) rules are constructed for predicting occurrence time of machinery failures in the future. The proposed rules provide explicit knowledge representation and semantic enrichment of failure prediction results, thus easing the understanding of the inferred knowledge. A case study on a semi-conductor manufacturing process is used to demonstrate our approach in detail.
Full PDF Version: 

Major Revision

Solicited Reviews:
Click to Expand/Collapse
Review #1
By Julius Mboli submitted on 06/Jul/2019
Minor Revision
Review Comment:

The paper is well written to the best of my knowledge. My only concern is the evaluation techniques used. The authors have used the online ontology pitfalls scanner called oops! to evaluate their novel proposed Manufacturing Predictive Maintenance Ontology(MPMO). The question I have is how reliable is this oops! and how can it guarantee that MPMO is reusable and salable since this is required of ontologist?
Secondly how can reviewers or even readers ascertain that the result of the evaluation as presented in figure 10 is actually for MPMO? This is because people can actually scanned the RDF of different ontology but report it as if theirs.
Thank you.

Review #2
Anonymous submitted on 16/Sep/2019
Major Revision
Review Comment:

This paper introduces a hybrid approach to facilitate predictive maintenance tasks in manufacturing processes. The proposal is a combination of data mining and semantic (using Manufacturing Predictive Maintenance Ontology and SWRL rules). A semi-conductor manufacturing process is used to demonstrate the presented approach.
The following is my suggestions. The authors may consider to revise the paper accordingly.
- The authors may highlight the motivation of this work. What motivates to propose this approach. A manufacturing process involves both hardware and software components. Where do the author see the limitation? An overall conceptual architecture may help reader to understand this interesting work.
- What are the limitations of applying Semantic Web technologies to Manufacturing process? The authors may present the technological difficulties/limitation faced by them.
- The authors may present alternate technologies of Semantic Web, which can be used to address the challenges presented in the paper.
- The authors may present the challenges in a separate subsection of Introduction Section. This would help reader.
- The authors may revise the introduction section, with a clear objectives with respect to the challenges.
- A table in the related work with a clear comparison of the existing approach is recommended. At the current stage, it is bit difficult for me to see a clear comparison among existing approach with respect to the proposed approach.
- In Section 5, the authors may mention the rational of choosing the technologies for the experiments.
- Section 5.1 may provide more information about the data fields of data set.
- If the public URL is available, then it should be included in the paper. So, the readers may refer the dataset for their experiments.
- If the work is open source, then public URL of the code repository should be included.
- Fig 9 may need revision. It is not readable. It must be presented as image (like Fig 10) into editors, mentioning the functionality of SWRL rules.
- There are major formatting issues in equations and tables. The authors are advised to fix them.

Review #3
Anonymous submitted on 01/Nov/2019
Minor Revision
Review Comment:

This paper proposes a predictive maintenance use case for Industry 4.0 use cases. The work is focused on combining data mining and semantic reasoning in order to improve the results of prediction and provide a better understanding of inferred knowledge. Outcomes of the data mining algorithms are used to build knowledge representing the prediction results which can be easily understood by novice users.

Paper is well-written and to some extent easy to follow. The authors have clearly laid out their contributions and structured the paper accordingly. Paper, as it stands, sounds like a decent contribution beyond the state of the art solutions. However, a few claims made by authors need further verification and justification. For example, authors claimed that having semantic representation will increase the readability and understanding for the novice users, such kind of claims need to be backed up by references. One can easily provide a counter-argument that even without using semantic representation, a better visualization technique can support a better understanding of trends and timely detection of future events.

Authors further need to motivate that how their approach of automatic creation of knowledge base and semantic representation is a better choice. A major advantage of semantics is the uniform information representation (data + metadata), which should have been demonstrated in this paper to show how the proposed approach is capable of providing semantic interoperability among various heterogeneous information sources.

The example given in the paper is a bit disconnected, thus making it hard to get a step by step picture of the overall flow. What about introducing a use case at the beginning of the paper by discussing the real-world scenario and portray the picture of how the existing system will fail and how the proposed solution will improve the status quo. I recommend authors to give an example scenario and then use this example throughout the paper.

Datasets and experimentation are also a bit vague and hard to show performance and value addition of the systems. I recommend improving the structure of Section 5 by properly introducing the dataset, aims of experiments and then discuss the achieved results.

Minor formatting issue: Figures/Tables are often on different pages than they are referred to in. I believe it is an automatic Latex setting, but perhaps some tricks can help to bring the figures/tables and text discussing them on the same page rather than turning between pages every now and then.

Overall, a decent contribution and a novel approach. Although hard to conceptualize overall achievement particularly in the practical scenario with real-world settings and usage by active workers within a smart factory.