Language on the job – From peashooters and packing fingers…

3. January 2020

Many manufacturers supply their processing and packaging machines to international customers. MADDOX is supposed to work for the same machine type in different languages. We are therefore often asked whether we can also automatically translate the knowledge it contains into other languages. The answer is yes: An interface to DeepL, currently the best translation tool based on machine learning, is high on our roadmap.

But what if there is another language in every company?

All plants have their own language, which is constantly developed by employees over time. Machine operators rarely use the technical terms for machine parts that designers have written down in the manuals. The latter are often found with the technology and are not memorized. Instead, an operator is usually trained by an experienced colleague who explains and shows him the machine. New employees have to invest a lot of time to learn this language. In the course of professional experience, the exact wording of technical terms can also be forgotten or modified by technicians. Terms that are too long are abbreviated. “Package holding finger” and “transfer finger” become “holding lever” and “transfer lever,”. One holding the package, the other moving it along. Unknown parts are given names according to their characteristic appearance or their main function. Sometimes very different parts are given the same or very similar titles, e.g. “Punch” and “Punch”. This can easily lead to confusion.

For a user survey in the factory of one of our pilot customers, we first worked through machine manuals. To compile the users’ understanding of the machines, we then had technicians, machine operators and shift supervisors explain them to us in detail. It quickly became apparent that users have different understandings of how a machine is built and how it works, and that there are many different terms for a component. For a component that blows air, we heard many terms, including “blower tube,” “blow tube,” and “fan rod.” Tape designation is also a good example. While technicians counted tapes by manual, namely from back to front, machine operators number the tapes as they usually stand in front of the machine. Thus, they give the number 1 to the belt closest to them and then number through to the back. Of course, this can lead to misunderstandings.

Findings from a survey with our pilot customers

Wir hatten eine außergewöhnlich spannende Messe und sehr viele interessante Gespräche. Während ein Großteil unserer Besucher unsere Idee als „nette, aber vollkommen realitätsferne Spielerei aus der Forschung“ abgetan hat, haben sich aus anderen Gesprächen langjährige, vertrauensvolle Partnerschaften entwickelt.

Zur Interpack 2020 (7. bis 13. Mai 2020) dürfen wir wieder auf dem Messestand des VDMA gemeinsam mit dem Fraunhofer IVV Dresden ausstellen. Wir freuen uns schon sehr darauf, an gleicher Stelle zeigen zu können, dass aus der „realitätsfernen Spielerei“ innerhalb von 3 Jahren ein Prototyp, ein Unternehmen und ein außergewöhnliches Team entstanden sind.

Is this a “punch” or a “punch”?

The technical language used in operations is complicated and often ambiguous.

MADDOX improves operational communication

In cooperation with engineering psychologists, we have developed various functions with which MADDOX not only enables exchanges for troubleshooting, but also improves the quality of communication:

>
To document faults and solutions, photos can be easily taken and annotated with text or finger sketches. It is also easy to record and insert video and audio recordings. This means that in many cases the often misleading verbal description can be omitted.

>
Our algorithms based on machine learning treat knowledge content as a “container” – regardless of the content. As a result, content that was previously difficult to search for, such as photos and videos, is found just as well as text. An error-prone and cumbersome keywording is not necessary.

>
We also use the concept of working as visually as possible for the localization of faults: The location of the fault can be graphically marked on a stored photo or diagram of the plant and thus documented unambiguously.

>
MADDOX presents stored knowledge in a structure and sequence that specifically promotes understanding and avoids misunderstandings. For example, in the event of a malfunction, a description of the malfunction is output first in order to create a common understanding of the problem and not to apply a correct solution to the wrong problem.