Can Machine Learning Help the Aerospace and Defense Industries?
Total world military expenditure rose to $1,739bn in 2017, an increase of 1.1% in real terms from 2016, according to new figures from the Stockholm International Peace Research Institute (SIPRI). With the increase in defence spending globally, one of the emerging global market trends within aerospace and defence is a renewed interest in the delivery of native language technical content to support foreign military sale (FMS) customers.
To understand this trend, let’s first define a couple of terms that drive leadership decisions within the aerospace and defence Industry.
Mean Time to Repair (MTTR) – The average time required to fix a failed component or device and return it to operational status.
Operational Readiness – The capability of a unit/ formation, ship, weapon system, or equipment to perform the missions or functions for which it is organised or designed.
These two factors contribute significantly to the success or failure of a military system. The take away? As MTTR decreases, there is a corresponding increase in operational readiness. In other words, a small decrease in MTTR could translate to one more piece of equipment being ready for the mission. This one additional piece of equipment could be the difference between the success or failure of that mission. For example:
“Maintainers accidentally fired the airplane’s Gatling gun and blew out their hearing in the process.” It’s not just a content quality or content accuracy problem…it’s a language problem!
There has always been the need for content quality and content accuracy, they’re major factors that help justify a modern content delivery strategy. After all, misunderstanding a procedure can be catastrophic... if one button is pressed in the wrong order, bad things can happen.
While the actual cause of the Belgian F-16 incident is still under investigation, it drives home how understanding could impact safety, MTTR, and operational readiness for foreign militaries. This accident could indicate a disregard, ignorance or absence of a procedure or process, or it could indicate a misunderstanding by the maintainers to some aspect of the instructions.
A Potted History of Controlled Language
Scientific studies from the 1970s to the present day – within the aerospace and defence industry prove the readability, consistency and usability of information directly affects a maintainer’s comprehension and efficiency. When managed well, these proven relationships positively influence MTTR, operational safety and operational readiness. English has long been the language of the aerospace and defence industries. In the 1970s, the aviation industry produced technical information in US English, British English and in native languages translated into one of these forms of English with sometimes wide variations on the resulting documentation.
Simplified Technical English (STE)
European aircraft manufacturers, challenged by this inconsistency, approached the Aerospace and Defence Industries Association of Europe (ASD) about the possibility to create a ‘controlled language’ approach to technical documentation. The ASD reached out to its American counterpart (Aerospace Industry Association of America or AIA) and together created a guide for simplified technical English that has evolved into today’s standard called ASD-STE100. Simplified technical English (STE) addresses a global aerospace requirement by aiding readability and comprehension for nonnative English speakers. STE is a way to ‘control language’. In simpler terms, it’s a boiled down set of English words that is used as a dictionary to aid in communicating core information.
This ability to convey complex methods, concepts and ideas without needing to translate or localise content has been a primary requirement for global aerospace companies. However, the times and market trends are changing and the value of native language content in terms of safety, MTTR and operational readiness is now becoming more recognised.
Use of a standard such as STE to control language is good but not appropriate for all projects. STE by itself will not objectively measure and report the quality of your content relative to your quality standards. It will not manage project, product, organisational or business level terminology and it won’t help you achieve high levels of content consistency or reuse. STE is a standard dictionary of words, nothing more and nothing less.
Why is there a need to simplify English? The English word ‘lift’ has 13 different meanings when used as a verb and eight different meanings when used as a noun. In another example, a MRO had an incident caused by ‘language’ resulting in engine damage in 2001. The English word ‘clean’ has two meanings: 1) get rid of paint, eg, strip, and 2) use cleaner to clean. The correct interpretation should be ‘get rid of paint; strip’ in this context. However, the mechanic did not understand, and performed cleaning by ‘use cleaner to clean,’ which resulted in wires burned from the cleaning fluid
Eliminate the Inefficiencies
The international language of the aviation industry, and all technical documentation is English. For 80% of aviation industry operational, maintenance, and support personnel, English is not their native language. Unfortunately, the use of STE alone to support militaries in various countries around the world has built-in inefficiencies. Non-native Englishspeaking maintainers must go through two years of English training before they start to learn how to maintain their assigned equipment.
The aviation industry is facing a shortage of maintenance technicians. Boeing has predicted a need for 679,000 technicians over the next 20 years with the majority of the need coming in the AsiaPacific region (39% or 268,000). According to the Aviation Technician Education Council, about 30% of the current ranks of aircraft mechanics are at or near retirement age and they’re retiring faster than they’re being replaced.
Dr Bill Johnson was recently asked, in his article Training and Job Aiding Technologies for Maintenance (September issue of the FAA newsletter), ‘aviation maintenance technicians (AMTs) and other industry personnel have the technical knowledge to safely perform their jobs, so why does failure to follow procedure (FFP) remain a leading cause in administrative actions?’ While Dr Johnson explores an organisation’s safety culture in the failure to follow procedure incidents, one has to wonder to what degree misunderstanding or incorrect interpretation plays a role in these incidents?
Without translated content:
Non-native English speaking technicians cannot perform their tasks at optimal efficiency.
MTTR goes UP!
Overall unit readiness is slower to reach optimal levels.
A shortage of qualified technicians to support the operational readiness of the system is created while the technicians are ‘in the queue’ for two years while learning English.
Use of simplified technical English helps reduce ambiguity but a technician must still have a degree of fluency to fully understand the details and intent of all instructions, directions and procedures.
Difficulty in hiring, training and retaining mechanics leads to higher programme costs, longer MTTR and decreased operational readiness. If we are aware of the negative impact on MTTR and operational readiness by not having native language content, then…
So why aren’t all FMS customers translating their technical data?
Because... It’s VERY difficult to do right and it costs A LOT of money to sustain a translation strategy.
Here are just a few of the sustainment challenges that impact translating highly complex, technical, military equipment data:
Very high volumes make traditional human translation strategies too slow and too expensive.
The content covers multiple areas of very complex systems, and some content covers highly complex and dangerous procedures that need to be translated with the highest accuracy.
Keeping translated manuals up-to-date is challenging.
Lack of security protocols and quality control.
What would it mean to FMS custors if there was a new way for original equipment manufacturers (OEMs) to deliver high-quality native language technical data quickly and efficiently? The benefits would be significant:
Lower costs to train and retain maintainer and technical staff.
Elimination of two years of English training before maintainer training begins.
Elimination of the shortage of maintainers due to short military service.
Higher accuracy and efficiency of maintenance procedures.
- Improved operational readiness.
Time to Abandon English?
So, should A&D manufacturers just abandon simplified technical English? No, not at all! The benefits of a common ‘controlled language’ have proven beneficial to the industry and STE is still a requirement for many programmes.
What I recommend is that all A&D manufacturers require a content quality systems initiative that objectively measures and reports on the quality of their data from the moment of content creation through sustainment, for the life of the programme. Proactively managing quality and consistency of terminology, writing style, content, reuse and tone increases the consistency of your source language content. This, in turn, facilitates an increase in language translation automation, which reduces localisation costs over the lifecycle of the programme. The more consistent your source content, the more efficient and less costly your localisation process can be, and the increase in content quality can save lives and improve operational readiness.
Speak One Language with Machine Translation
Today, machine-based translation has matured at such a rapid rate that both the speed and accuracy is now able to deliver a more economically viable approach to delivering technical content in almost any of the languages required by global aviation maintainers. Machine translation (MT) is the translation of text by a computer with no human involvement.
Pioneered in the 1950s, MT is also often referred to as automated translation, automatic translation or instant translation. It is estimated that as many as 600bn words a day are translated by computers today, across the various MT portals. This dwarfs what the localisation and professional business translation industry does by a factor of more than x99.
Machine translation has evolved from simple rules-based machine translation (RBMT) to statistical-based machine translation (SMT) approaches to today’s neural machine translation learning (NMT) network approach. Neural MT systems also utilise machine learning approaches but these systems learn higher-level concepts for producing translations. In this new approach, the system that is created is a multi-layered neural network that produces translation in a similar processing pathway that the human brain would follow.
The neural technologies have become increasingly popular among MT researchers and developers because NMT systems have shown to produce translation with a fluency that RBMT and previous SMT systems could not rival, especially when the sentence contains more complex information structures. Neural MT uses a deep learning architecture capable of learning the meaning of the text which enables the machine to perform the translation task at a semantic level leading to fluent and naturally sounding translation output.
Maintenance errors can prove costly for both the aircraft and human life. Simple misunderstandings can have devastating impacts or just prolong the time it takes to restore an asset to operational readiness.
The aerospace and defence industry is facing a global shortage of aviation maintenance technicians currently and well into the future. While English is the language of aviation and of aviation maintenance, 80% of maintainers speak English as a second (or even tertiary) language. English has a role to play as the ‘control’ language but local or native language can aid in training efficiency and in understanding by maintainers. Local language can facilitate operational readiness by helping to fill the need for new highly-skilled technicians more quickly.
New advances in translation technologies have increased the speed and accuracy of machine translations and dramatically lowered the cost such that it is now more economically feasible to deliver the most technical of content across languages.
Operational leaders must consider new thoughts and approaches to better enable maintainers through the technical documentation and training that is created and delivered. To that end, as leaders in the A&D industry, we must explore strategies that overcome the high cost and long lead times for language translation that has historically prevented the consideration of multilingual technical content as a viable consideration to help increase maintainer understanding, improve MTTR and operational readiness.
View original content here