When lift inspectors talk to themselves

The first lift inspectors at TÜV Nord now have the support of AI (artificial intelligence) during their work. From now on, the specialists will document defects during inspections by voice entry and no longer as before on the computer.

This should save them time and have the advantage of even greater precision.

"Previously, the experts had to type the defects observed into a field in a computer. Now they conduct a conversation with AI via their smartphones," revealed Daniel Contreras Schaffeld, business development manager at TÜV Nord and responsible for the implementation of AI-supported voice recognition software for lift inspections. "To an outsider, this may sound like they’re talking to themselves but it greatly simplifies the work, since their hands are free and they can concentrate completely on the inspection steps."

The AI now accomplishes the documentation of the inspection, i.e. the assignment of the identified defects to categories. After the inspection, the experts can concentrate on the quality assurance and their report since the identified defects have already been recorded.

Long training phase

The introduction of the AI-supported voice recognition software "voize" was preceded by a long training phase. First, the relevant defects and frequently occurring indications for describing defects had to be recorded. The AI then had to be trained to recognise, interpret and correctly assign the spoken terms and defect classifications.

The advantage for the experts is that they no longer need to learn a term catalogue by heart. Moreover, the AI now also understands the terms if they have in the meantime changed, such as the term "isolation valve", which is now defined as "line break valve". In addition, the AI is trained to correctly interpret its "own" terms and the abbreviations of the experts.

AI must understand dialect

"However, it is not enough that KI learns these terms," according to Daniel Contreras Schaffeld. "People are people and not machines." The AI has to understand and correctly react to dialects and linguistic variations. For this reason, the AI was tested with twenty experts from throughout Germany and in this way trained on voices and dialects.

In future, the software is also to be deployed in other areas; these include for example escalator inspections. Moreover, it is also to be offered in other languages to support employees of TÜV Nord in other countries.

More information: tuev-nord.de/de