Towards Welding Knowledge Models as ISO-Driven and Standardized Ontologies

Tracking #: 3048-4262

Muhammad Yahya
Baifan Zhou
John Breslin
Muhammad Intizar Ali
Evgeny Kharlamov1

Responsible editor: 
Guest Editors SW for Industrial Engineering 2022

Submission type: 
Full Paper
The ongoing Fourth Industrial Revolution (I4.0) witnesses a series of profound changes in the industrial communities such as automotive, aerospace, manufacturing, etc. Widely applied in almost all these domains, the automated welding process is an impactful manufacturing process that shifts towards intelligent welding systems that take advantage of emerging technologies of I4.0. Welding refers to a complex process that joins together materials. It is carried under certain standard rules and guidelines defined by various international consortiums such as the International Organization for Standardization (ISO), the International Society of Automation (ISA), and others. The current knowledge engineering practice of welding is done in a fashion that follows expert heuristics, rather than standards, generic ontologies, etc., possibly due to a lack of generalizable, commonly agreed, and high-quality welding ontological models. In this way, many industrial ontologies are created based on specific welding heuristics or datasets. They cannot be guaranteed to be capable of fully capturing welding domain knowledge, to pass ontology tests checking metrics such as clarity, consistency, completeness, etc., and to be generalized and reused in other welding applications or similar processes. In this paper, we propose our modular generic Welding Core Ontology (WCO), to formalize the welding domain knowledge. WCO includes modules of Business ontology, Physical Entities ontology, Process ontology, and Product ontology. It is built on the knowledge obtained after extensive discussions with Bosch welding experts, analysis of good coverage of standards by ISO, ISA, etc., considerable reuse of the Reference Generalized Ontological Model framework, and reasonable alignment with existing established ontologies. WCO is evaluated under four dimensions of requirements: knowledge capturing, quality ontology, conformance to standards, and conformance to existing ontologies. The evaluation shows that it reasonably satisfies all of the four requirements. The evaluation combined with lessons learned reveals the great potential of WCO for industrial as well as academic communities.
Full PDF Version: 


Solicited Reviews:
Click to Expand/Collapse
Review #1
By Bob Young submitted on 16/Mar/2022
Review Comment:

General comments:

This paper has some originality from a spot welding ontology perspective, but the results are not really significant in the context of their claims. The quality of the writing from a grammatical perspective is generally good although there is some significant repetition through the article. However, the academic rigour in arguing their research contribution is poor.

It is clear that the authors have undertaken a substantial piece of work, and it is good to see that they have been working closely with industry experts. However, the paper lacks rigour in much of the writing, with the claims that they make being well beyond what they have achieved. In particular they claim:
(i) to have produced a generic welding core ontology whereas they have really focused in detail on only one of many welding processes; resistance spot welding. Resistance spot welding is one of 8 categories of resistance welding, according to ISO 4063 which they use a major source, with a further 4 main categories of welding all with multiple sub-categories. Their work on that basis is well short of providing a generic welding core ontology.
(ii) They also claim to have analysed a ‘good coverage of standards’ while what they appear to have done is picked those standards that suit their needs. Whilst this approach might just be acceptable in a very narrow domain-based activity, they can hardly argue that they have produced an ‘ISO-Driven’ result without reference to other related standards. They do not reference the many existing ISO ontology, product and process related standards, or position themselves in relation to the major ontology standardisation initiatives currently underway.

On this basis I suggest that the authors reconsider what their true contribution really has been and submit a paper to make that argument, without radically overstating their achievement.

Other specific comments:

1. Introduction:
The first paragraph of the introduction is concerning as it does not provide a clear or correct understanding of the relationship between joining processes, welding processes and the resistance spot welding process, or a clear understanding of the relative importance of resistance spot welding to industry.

The second paragraph has a focus on standardisation, which is good, and that many standardisation bodies have welding standards. However, no view is expressed as to whether these standards are consistent with each other and, if not, what that implies in relation to their intention of “standardised documents can be used to semantically represent the welding domain to build an intelligent system”. This lack of consistency is a critical issue that needs to be addressed along with the level of intelligence, or lack of it, that is built into standards.

2. Related Work:
Given your interest in standardisation approaches and ontologies, I would expect to see references to the work of the Industry Ontology Foundry and the work of OntoCommons in your related work section, along with an explanation of how you position your work in relation to these activities. I note that you mention these initiatives towards the end of the paper, so given your awareness of these you should make it clear how your work should have value within their contexts.

3. Requirements for Welding Ontology:
In your requirement to capture domain knowledge you state that “capturing domain knowledge in order to build an ontology is a first requirement”. Whilst that may be arguable, my main concern here is how do you manage this across multiple businesses with different terminologies and different specific detailed knowledge that relates to their business? It is good that you have worked closely with an industry user of the technology, but only one user. You need to reflect on the limitations that this brings to your results.

There are similar issues with conformance to ISO standards and existing ontologies. This can work when you choose specific standards and specific ontologies but cannot be achieved in general across the existing range of standards and ontologies. These issues are not clearly addressed in the paper. Although you do provide some insight in your section 5 on welding standards, you do not clarify the limitations on your approach to building a generic welding ontology when you do not address the differences across standards.

4. Ontology development methodology
In section 4.1 you mention that you have undertaken a comparative study with ISO 4063, ISA-95 and RGOM. You should explain why you have chosen these and more importantly why you have discarded the many other potentially related standards and ontologies. A minor additional point, given the importance of international standards, you might want to reference IEC/ISO 62264 as the international standard based on ISA 95.

6. Welding Ontology
Para 2:
To say “we adopt the RGOM framework and industrial standards along with other existing LoD vocabulary” is unhelpful as it is too general. Which industrial standards do you adopt and which LoD vocabulary is important to you? Also why?
The last sentence in this paragraph is not clear.
Para 3:
It would help if you justified the choices of the modular ontologies you mention. It seems strange to have high level business and product ontologies alongside a relatively low-level welding process ontology and a very general physical entity ontology.
Para 4:
Your sentence “ISA-95 represents the business entities layer by layer via layer 0 to layer 4 from enterprise to production process” is incorrect and confusing. Business entities in ISA-95 are represented at level 4. If you are to argue successfully for an effective, useful ontology it is really important that you are accurate in your descriptions. You need to explain correctly the levels of ISA-95 and how they are used in your work.
Also you state that “ISO 4063 is used to model the processes and products ontologies”. While ISO 4063 mentions many processes it does not refer to products. So, what product ontology is your work related to? There has been a great deal of work in this area that you do not address.

6.1 Business ontology
You state that “The concepts and relations in the business ontology represents
the semantics being adopted from ISA-95”. This does not appear to be true. Figure 4 does not represent the levels defined in ISA 95. You may have based your figure on reference 32. If so, that is what you should state and not that what you have done is based on ISA-95. You need to clarify how your overview fits with ISA-95. Also the left half of figure 3 is not consistent with ISA-95 levels. To summarise this issue it is not clear that what you have done is consistent with ISA-95 or in what way it relates to ISA-95.

6.2 Physical Entities Ontology
The breakdown shown in figure 5 may show the main physical entities involved in a welding process but it is inconsistent with the physical entities column shown in figure 3. This issue needs to be addressed.

6.3 Process ontology
You state that “Welding is an assembling process where two or more components are connected together to form a joint. The joint is formed by the arc of an electrode which used some power source, pressure, or both”. This is not correct and needs to be restated clearly, concisely and correctly.

Figure 8 is not referenced in the text.

In figure 10 it would be helpful to illustrate what is taken from RGOM and what is 'new' or taken from other ontologies. The difference between ‘process and ‘operation’ should also be explained.

6.4 Product Ontology
To use the notion that a welding spot is a product is quite unusual. As mentioned in relation to your claim in section 6, this does not fit with ISO 4063. ISA-95 is closer to what you describe in ‘the desired output of the processes of an enterprise”, but this implies a combination of processes with an output from the enterprise, so hardly an individual welding spot. You need to justify this notion, perhaps in the context of a manufacturing process, but if someone was buying a product from Bosch or any other supplier of goods that included welding spots, they would not consider a welding spot as a product.

8 Evaluation

8.1 competency questions:
The competency question 2 in table 5,SPARQL query: this is a very specific definition of a welding facility. There are a range of other types e.g. you do not need to have a robot to undertake welding; also there are other types of welding. As such it may be appropriate for Bosch’s interests but will not be ‘standardised’ across multiple businesses.

In the competency question 4 the SPARQL query is not specific to the welding process, but does require a definition for ‘operation’.

You state in the text that” In our case, the competency questions were provided by the Bosch experts, and thus it was not required to analyze the efficacy of the competency questions”. This raises 2 issues: (i) have you checked that the queries provide the answers that the experts expect? (ii) what have you done to explore potential competency questions from other industrial companies?

You state that you have executed a SPARQL query and show the query. What you do not mention is how many other queries were executed. If only one query then to what extent can you say that you have validated the work you have done?

8.3 Generalisability

You state that “The welding process and physical entity module of the ontology can be easily adapted to a wide range of welding processes”. It cannot be adapted to all welding process but you should state which processes you believe it can be adapted to and why.

8.4 Reusability
Your reusability claim across industries is really saying if everyone agrees to use your ontology then it is reusable, but that is true of all ontologies and the problem with most manufacturing ontology research is that the results are only very narrowly usable. You need to address how your work can contribute to reusability, given the wide range of similar but distinct industry requirements.

9 Lessons learnt
You mention that discussion with domain experts is difficult although vitally important. This is very true, but domain experts from different industries will have variations in their expertise and, as in the comment above, you need to address how your work can contribute effectively to solving this problem in an open and flexible way.

It is good to see you now refer to the OntoCommons and IOF initiatives in this section, but for your work to have real value to businesses outside of Bosch it must be possible to understand how your work should be positioned in relationship to these initiatives. This is really important if you are to argue your contribution in the context of standardised ontologies.

Review #2
Anonymous submitted on 27/Apr/2022
Review Comment:

Disclaimer: I would like to start by noting that the use of the English language in this paper is suboptimal (several and severe grammatical errors, unusual constructs, unclear formulations) which often makes the text difficult to understand. While I read the paper in detail, there have been several passages that I could not fully understand and therefore this might have an impact on the accuracy of the review.

The paper describes the design of the Welding Core Ontology (WCO) and its evaluation. This is an important problem, because welding is a process that is used in several industries. Despite the importance of the welding process, no welding ontologies exist. Currently, a number of standards for welding terminology exist but are used in different geographical regions. The authors set out to fill such gaps in the area by the design of the WCO which addresses a set of requirements.

As this manuscript was submitted as 'full paper', it should be reviewed along the usual dimensions for research contributions as follows:

(1) Originality: Low. If considered as a full paper (regular research paper), this paper lacks originality in terms of the addressed problem (ontology construction in industry, in a domain where already several ontologies exist); approach (classical approach for modelling and evaluation are used); and output (it is not clear in what ways the resulting ontology is novel with respect to other industry ontologies).

(2) Significance of the results: Low. Currently, the paper insufficiently proves that the created ontology will have a major impact on the area. In fact, as stated on page 12, this ontology has not yet been used in an “enterprise production line”. Since the paper follows classic steps for ontology design and evaluation, there is also limited significance at the methodological level.

(3) Quality of writing: Very Low. The major issue of the paper is the suboptimal command of the language. Some of the used expressions are rather colloquial (“should be around”, “it is hard to come up¬¬”), there are countless (very basic) grammar mistakes, and some sentences are formulated in ways that are ambiguous and difficult to understand. Therefore, the paper requires a major writing level revision to bring it to the standard expected for publication in a journal.

The considerations above are based on the following more detailed comments, which I hope will help the authors to improve their work.

Related to the requirements: it is unclear how these were derived. Which stakeholders/ goals lead to such requirements? Also, requirements R1 mandating that the ontology captures domain knowledge seams to contradict point 2 at the end of Section 2 stating that a major gap in the literature is the fact that existing ontologies “are developed from the perspective of domain knowledge”.

Related to the ontology development methodology, it would be interesting to know whether other methodologies were also considered and why the one of Falbo was selected. Also, the methodology description remains imprecise in several aspects: e.g., how many meetings were held with how many experts? What was their background? What method was used (interviews, workshops, focus group)? “In this regard, a comparative study is conducted with ISO 4063, ISA-95, RGOM and other existing vocabulary to analyze the concepts being covered.” – where is this comparison reported (is this section 5?)? What was the main conclusion?

In terms of reuse, the authors focus on reuse at vocabulary level (e.g., isPartOf from DC). However, modelling of meronymy is a rather complex topic and there exist several ontology design patterns for modelling such information. Have the authors considered the reuse of such recommended modelling practices?

The designed ontology is sufficiently well-described. It could be improved by: showing picture 13 earlier so that the reader can identify the key modules and their sources early on. Several parts of the ontology require representing entities that measure some physical quantities, but the choice of ontologies for representing units of measurements is not discussed in text (only hinted in Table 1) – consider dedicating some discussion to this aspect.

It is commendable that the authors have tried to evaluate the ontology from several perspectives. However, much of the evaluation section describes background theory while often remaining unprecise on the employed methodology and the obtained results.

The evaluation by the two experts would be the most interesting, however, several details are missing about how exactly this evaluation was performed (who were the experts? How did they interact with the ontology? How did they provide the feedback?) Also, if these were the same experts that provided the domain knowledge as input to ontology construction they might have been biased in what they expected. “Finally, the ontology alignment with existing vocabularies is evaluated from the ontology experts.” – how was this performed? What was the conclusion?

The competency question based evaluation is limited to 5 queries. The reusability section (8.4) actually focuses on the fact that WCO makes use of other ontologies, as opposed to demonstrating that WCO can be reused in several settings domains.

Other smaller comments:
(1) As this is a Semantic Web journal, the introductory text for what the Semantic Web ais and the introduction to ontologies is not strictly needed for the audience of this journal. Reference 13 is not representative for introducing the notion of ontologies. Reference 4 is about ontology learning and is not a suitable reference for the use of ontologies in I4.0.
(2) Section 5 describes a number of standards in the welding area, however it has little value for the paper as only the ISO standard is then finally reused in the ontology. Therefore this section should be reduced to focus on the information that is relevant for the paper.
(3) Some parts of section 5 that clearly exceed the rest of the paper in terms of text quality are copied from online sources [1] – which could be seen as plagiarism of content.
(4) Section 6.1 – different spellings are used for the terms in Fig 4 and the text describing the figure.
(5) References are not correctly capitalised, which is especially problematic for references that contain acronyms such as SSN (in [10]) or proper nouns (Bosch in [17].


Review #3
Anonymous submitted on 06/May/2022
Major Revision
Review Comment:

This manuscript was submitted as 'full paper' and should be reviewed along the usual dimensions for research contributions which include (1) originality, (2) significance of the results, and (3) quality of writing. Please also assess the data file provided by the authors under “Long-term stable URL for resources”. In particular, assess (A) whether the data file is well organized and in particular contains a README file which makes it easy for you to assess the data, (B) whether the provided resources appear to be complete for replication of experiments, and if not, why, (C) whether the chosen repository, if it is not GitHub, Figshare or Zenodo, is appropriate for long-term repository discoverability, and (4) whether the provided data artifacts are complete. Please refer to the reviewer instructions and the FAQ for further information.

(1) originality

The paper describes "welding ontology" developed by the authors and the associated development process. While it seems there is not many welding ontology out there, the described ontology still is more like an information model rather than ontology. This is because there is no defined axioms. The authors discussed the constraints and with the OWL language the reasoner does not validate against those constraints b/d OWL reasoners are open world.

(2) significance of the results
In the above sense, the so call ontology is of limited significance. The value is more on the knowledge translated from standards to be represented in a more computer processable knowledge graph. But there is no pointer to the knowledge graph to be reused by the community.

Ontologically, the authors seem to be less aware of semantics at the fundamental level. For example, the ontology is said to be modularized into physical entities module and product module, among other. How is the product not physical entity? And how is Factory Area or Production Line not a physical entity?

(3) quality of writing

The paper suffered the most of the writing quality. From the Abstract section down to about section 3, I can find spelling or obvious grammatical error almost every the other sentence. For example, the author mixed up prospective vs. perspective, "sensible manufacturing resources" - what is sensible manufacturing resources, this is just two examples from numerous errors that I don't have time to list all.

In conclusion, I'm leaning toward rejecting the paper, but gave a major revision because I think there was a good amount of work went into producing the welding knowledge graph. But that knowledge graph (with individuals) should be made available.