Terminology study on vibration-based condition monitoring technique

1. Introduction

Technical terms are an essential part of both technical and scientific authoring. Engineers and experts use a vocabulary that relays a variety of specialized concepts by means of technical language. This special vocabulary conveys concentrated meaning that has been built up over the years in a specific domain. The value of terminologies – the lexical components of specialized languages - lies in the way each concept condenses a mass of information into a single-word or a multi-word expression. In our research, we introduce a terminology study on the application of a vibration-based condition monitoring (CM) technique, which is a keytool in the predictive maintenance of any equipment and machines within a wide range of industrial applications such as transport, oil and gas, processing and manufacturing.

CM provides a myriad of benefits, including 24×7 remote monitoring, early warnings of potentials (serious) failures, reduction in unplanned downtime, reduction in maintenance costs, support ongoing reliability and risk reduction with fewer inspections. CM uses measurement of specific equipment parameters, such as vibrations in a machine, its temperature or condition of its oil, taking note of any significant changes that could be indicative of an impending failure. Vibration-based CM refers to the use of in site non-destructive sensing and analysis of system characteristics in the time, frequency or modal domains for the purpose of detecting change, which may indicate damage or degradation.

The area of vibration analysis is filled with different technical terms, jargon, acronyms and concepts. General words that describe some general or vague group or class such as frequency, order, node, etc. lead to semantic confusion in the area of vibration. In fact, vocabularies applied to specific conditions are used interchangeably to denote general dysfunction and vice versa. This confusion may reflect research, which has radically changed over the last decades. To lessen this confusion we aim to provide firm and unequivocal use of terminology of vibration analysis.

While terminology identification, i.e. creation of a list with domain specific vocabulary for a textual corpus, has become widespread nowadays, the notion itself of term is still not clear, both from a pure linguistic and a computational point of view [1]. Juan C. Sager [2] defines a term as a depository of knowledge and a unit with specific reference in that it “refers to discrete conceptual entities, properties, activities or relations, which constitute the knowledge space of a particular subject field”. Similarly, Christian Jacquemin [3] and Maria Teresa Pazienza [4] define a term as “a surface representation of a specific domain concept”.

Quite a bit of work has been done in order to identify terms for a specific domain. Most of them use statistical methods to export terminologies from textual corpora, i.e. collection of documents in the domain. Many of the results are entirely usable. Gerard Salton [5] suggested term frequency-inverse document frequency (TF-IDF). It is a statistical measure that evaluates how relevant a term is to a document in a collection of documents. This is done by taking into consideration: (i) number of times a term appears in a document, and (ii) the inverse document frequency of the term across a set of documents. Later, Kenneth Church and Patrick Hanks [6] proposed the mutual information (MI) which is a concept rooted in information theory. In particular, it measures how much information is communicated, on average, in one random variable about another. Both variables are sampled simultaneously. Ted Dunning [7] proposed a measure, which is based on likelihood ratios and relies on frequency profiling. The method delivers reasonably results and works in both large and small text samples. Recently, researchers have used corpora comparison to capture n-grams (i.e., occurrences of one or more words) based on normalized frequency. For instance, Adam Kilgarrif [8] experiments with several measures, among others χ2-test, Mann-Whitney rank, t-test, MI, and TF-IDF and concludes that χ2-test performs best. In general, though, most methods lead to endless lists of candidate terms.

In this work, we present the initial results of a terminology work which is still in progress. We created a termbase, namely a collection of electronic term records from the vibration-based CM area containing 20 typical terms. Each entry, which is automatically extracted by domain specific textual corpora is validated by a team of subject matter.

2. Methodology

In this research, we extend the methodology proposed by Patrick Drouin [9], where term extraction is performed in domain-specific corpora leveraging information from general corpora. We are applying TF-IDF as well as syntactic information and stop-word lists to reduce the noise in the list of candidate terms by restricting terms that are sub-terms of other terms.

Our algorithm identifies both single-word and multi-word terms and it focuses on extraction on either verb or noun phrases. For verbs, it excludes auxiliary and light verbs and takes their base forms as candidates. For noun phrases, it extracts structures of noun-noun (NN), adjective-nouns (AN) and its combinations, including conjunctions (C) or determiners (D). It also pays attention to unknown words (e.g., words of the domain-specific corpus which are not found in the general corpus) because most technical jargon is not likely to be included in a non-specific dictionary.

3. Results

The list with the candidate terms consists of approx. 20,000 English entries, after the automatic validation. Table 2 shows the top 40 terms sorted by frequency of their occurrence. The following list does not include units, acronyms, abbreviations and multi-word terms such as vibration signal, vibration analysis, etc. Some examples of candidate terms among their structure are shown in Table 3.

This list contains lexical items that reached the probability threshold the algorithm used during the identification process. Further manual cleansing is needed to reduce the size of the list. For instance, the candidate term study which occurs 360 times will probably be removed from the list if it is not a representative term of the domain of vibration analysis. This is normal taking into consideration that our corpus consists of scientific papers where standard phraseology is used by authors. Hence, all candidate terms with high number of frequency such as example, addition, etc. need further validation for their representativeness in the domain by experts in the vibration-based CM domain.

The validation is done taking into consideration: (i) the representativeness of the term in the domain, (ii) if the term is neutral and free of judgements, (iii) if the term has been deprecated, and, (iv) if the term is unambiguous. Each term is accompanied by metadata; namely, list of contexts, POS, abbreviations, frequency, etc. (see Fig. 1). The experts could either Approve or Reject the candidates based on their knowledge in the area. In addition, they revise the terms by assigning definitions and more metadata or notes, if any.

Fig. 1Metadata displayed for the cacidate term acoustic emission

In this research, we present our findings from the preliminary results. In more detail, we present 20 ambiguous but typical terms that may lead to semantic confusion in the domain of vibration-based CM. These terms are employed in various disciplines including music, physics, acoustics, electronic power transmission, radio technology, and other domains, as well as they have a non-specific meaning. For instance, among others, the term “node” (http://www.electropedia.org/iev/iev.nsf/6d6bdd8667c378f7c12581fa003d80e7?OpenForm&Seq=2), according to IEC’s electropedia is used in the domains of acoustics and electroacoustics, circuit theory, mathematics, telecommunication networks, teletraffic and operation, electric traction. Similarly, the term “filter” (http://www.electropedia.org/iev/iev.nsf/SearchView ?SearchView&Query=field+SearchFields+contains+filter+and+field+Language=en&SearchOrder=4&SearchMax=0) is used in the domains of radiology and radiological physics, oscillations, signals and related devices, electrical and magnetic devices, lighting, etc. It is important to point out that a term from one domain may be borrowed and attributed to a new concept in another domain within the same language. Hence, the meaning of a term may conflict with the general language meaning, or it may confuse the multiplicity of technical meanings.

Under the light of the above, a team of experts in the domain of vibration-based CM revised 20 ambiguous terms (Accessible at: rb.gy/ieb7nh). It is part of the warm-up process of the term validation task given that they don’t have previous experience in terminography. For that reason, 2 hours of training was offered aimed to introduce the experts in the terminography. It should be noted that we recognize and acknowledge the valuable terminology work made by the International Organization for Standardization in conjunction with the International Electrotechnical Commission (IEC) [15] as well as by the Vibration Institute (https://www.vi-institute.org/vibration-terminology-project). In our work, we borrow definitions from both contributions and we expand metadata for the terms.

Terms are classified according to their relation to the concepts taking into consideration the principles of terminography; (i) synonyms are grouped together (in the same entry) and (ii) polysemes (i.e., words with the same pronunciation and/or spelling) and homonyms (i.e., (homographs and/or homophones) are presented separately (different entries) because they represent different concepts. Moreover, the term formation shall be concise and as neutral as possible, avoiding connotations, especially negative ones [16].

The terms presented in this work are based on the English language and are not intended to cover all the methods used for English term formation. Fig. 2 shows the term entry “harmonic” in XML format (According to [17]). In more detail, in the subject domain of vibration, the English term “harmonic” was formed as a simple single-word term, via terminologization of the ordinary term “harmonic”, meaning “repeating signals, such as sinusoidal waves” in order to render the concept “harmonic vibration, the frequency of which is an integral multiple of the fundamental frequency”. It is a singular, masculine noun and its deprecated term is “overtone” because “the term “overtone” has frequently been used in place of harmonic, the $n$th harmonic being called the ($n-1$)th overtone.” An example of usage of this term in the domain of vibration is “Some of these harmonics have a dominant value in the vibration spectrum due to interaction of machine flu harmonics and the mechanical structure of the machine.”

Fig. 2Metadata displayed for the cacidate term “harmonic”

4. Conclusions

We introduced a terminology study on vibration-based CM technique aiming to harmonize current definitions and terminologies. To do so, we automatically exported candidate terms from scientific research papers and as well as websites specialized in this domain. In this study we present the findings from the first results which are manually validated by a team of subject matter exerts. In more detail, 20 ambiguous but typical terms that may lead to semantic confusion in the domain of vibration-based CM are handled and documented. Terminologies are compiled in XLM format so that can be read by machines. This choice of the terms provided for the possibility to expand in the near future the temrbase rendering successfully multi-word terms that consist of them. For instance, by documenting the term “frequency”, the terms “frequency spectrum”, “rotational frequency”, “double frequency”, will be documented with less effort. The findings of his work will not only help scientists in the domain of vibration-based CM, but also teachers, professors and students who study this domain, as well as professional communication mediators, such as journalists, technical authors and translators [18]. It is also worth noting that this work may help language engineers to solve natural language processing (NLP) tasks such as domain adaptation of machine translation, text classification, search engine optimization, language understanding, etc.