You are not alone, IBM and Watson. Natural language processing (NLP) and machine learning (ML) are coming to the shop floor in a variety of ways and places. For example, Sydney-based 1Ansah is using both to help mechanics easily use technical documents while working on helicopters at Airbus Group Australia Pacific.

NLP ingests digital maintenance manuals, component manuals, service bulletins, airworthiness directives and fault-isolation manuals, then analyzes sentences and grammar, explains 1Ansah CEO Anant Sahay. It then creates an index specific to maintenance tasks. ML then uses statistics to learn from troubleshooting successes and exploit that experience in advising technicians confronting similar problems in the future.

Sahay says his tool, called AMRIT, has been fully deployed. An Intelligent document search engine for product life cycle management, tapping about 1.2 million documents, and Airbus’s business policies and procedures, including 10,000 documents, is now being used heavily by both mechanics and engineers. “The fault isolation system has been deployed for one model of helicopter and is under trial, while training is done,” he says.

Hardware consists of a Linux server farm with seven servers, all in Airbus Helicopter’s infrastructure. AMRIT uses replication and parallel distribution to execute the data products in a fault-tolerant mode using multiple servers. “This is to cope with high volumes of queries and data,” Sahay says.

Two other steps remain to be taken. AMRIT must ingest documents for more helicopter models and for fixed-wing aircraft. Then training and testing must be done. Sahay expects the whole project to be completed in the third quarter of 2017.

The Airbus project will be just the beginning of 1Ansah’s support of aircraft maintenance, according to Sahay’s plans. He is now discussing the solution with SIA Engineering Co., Emirates Engineering, Etihad Airways Engineering, Boeing Australia and Lufthansa Technik.

Sahay says the successful experience with Airbus Helicopters will ease future applications to some extent. “Implementing the core system will not be very difficult. However, most of the implementation efforts will be dependent on the number of types of aircraft and the format of the documents, that is, XML, SGML, PDF and so forth.” In any case, the helicopter implementation will save time and effort in the future. “It helped fine-tune the AMRIT platform itself and creates a framework for development and deployment,” he adds.