SparkCognition builds leading AI solutions to advance the most important interests of society. The company helps customers analyze complex data, empower decision making, and transform human and industrial productivity with award-winning ML technology and expert teams focused on defense, IIoT, and finance.
In case of commercial flights, to maintain their status and punctuality is the highest priority. In fact, even a single delay of one or two hours in flight can cost $10K or $150K per instance and subsequently damages the reputation of the airline. When it comes to military operations, keeping the flights at peak mission-readiness every single minute is extremely crucial.
Although extensive sensor coverage is established in modern aircraft but air to ground communications systems can only record simple telegraphic information or any fault, along with anecdotal observations from the flight crew. As an aircraft is built with millions of components, even an experienced professional would take a long time to spot the issue and eliminate it. This gives rise to customer dissatisfaction and increases the delay in flights.
Additionally, the demand for trained personnel will comparatively increase from the supply by 9 percent in the coming decade and the knowledge imparted from retiring technicians is often lost, with no significance to future workers.
DeepNLP creates a system that prescribes optimal maintenance solutions by mining historical maintenance records
To curb this most common issue with planes, a plane manufacturer was looking for a system that can reduce time on maintenance incidents and capture tribal knowledge of veteran employees. The company collaborated with SparkCognition™ for a solution. The solution commenced with the process of sorting through historical maintenance logs with an objective to minimize scheduled downtime by supplying front-line maintainers with most effective corrective action procedures for fixing problems.
Using its natural language processing platform DeepNLP, SparkCognition mined 120,000 work orders from 18 years to identify faults, successful corrective actions, and references to manuals that explain the proper procedures.
Working of DeepNLP
It is an ML-powered NLP platform which automates the retrieval of information, classification of documents, and content analytics. Pulling from unstructured, natural language content, such as written documents or images, and transform the information contained in that content into structured data, such as tables or categories, is its core functionality.
The project took around 90 days to complete and served the client with normalize variability in natural language to identify maintenance actions, inferring missing information from post-hoc and incomplete maintenance logs and identify a subset of records with clear actions that successfully resolved the problem.