Strategic collaboration on machine learning

Strategic collaboration on machine learning

Dipai is a newly established IT company that emerged from the innovation environment at Norwegian University of Science and Technology (NTNU). The company’s expertise lies in data analysis based on artificial intelligence and machine learning. Digitread IoT and Dipai have entered into an exciting strategic, joint collaboration.

The company name Dipai is short for “Diagnosis, Prognosis using AI“, and this is precisely the core of Dipai’s business idea. Their work focuses on developing systems that enable predictive maintenance to take place.

In simple terms, the method involves collecting data from different types of components over a period of time. The purpose is to find the normal operating pattern of components such as pumps or motors. The data collected from the components is used to train the machine learning algorithm. When the algorithm “knows” what is normal, it also finds any deviations from the norm. It is these deviations that give an indication that something is about to degrade, be destroyed, or malfunction which could lead to a breakdown.

“If we can give early warning, or predict deviations that could lead to a breakdown, we can send out service personnel prior to the components actually breaking down. Being able to detect such deviations, also minimizes the downtime of the components“, states André Listou Ellefsen who is CEO of Dipai.

“We start with data collection and spend some time training the algorithm, which, in turn, will provide the means for diagnoses to be made whereby we can predict with 98 percent probability that a specific malfunction or outage will or is about to occur. This will better facilitate forecasting, for example, where it predicts a component will break down in two weeks’ time, and that is the position we want to get to, so that the maintenance can be done before an outage physically occurs,“ he continues.

Inspection of real-time data on the fuel consumption of the research vessel Gunnerus is compared with the predicted value from DIPAI’s machine learning algorithm. Lars Hvidsten and Christoffer Lange study the application developed by Digitread IoT.

Parties that complement each other
“For Digitread IoT, machine learning is not a buzzword, but something we can actually realize. Machine learning is a tool the industry should use to reap the benefits of its data. Building systems with machine learning requires extensive knowledge, and we are extremely pleased that, together with Dipai, we can now offer the best expertise in the field,“ confirms Christoffer Lange from Digitread IoT.

While Dipai is a master in analysis and deep machine learning, Digitread IoT sits on the expertise and solutions around the other building blocks in the IoT solution.

“We have now created an IoT solution around Dipai’s machine learning whereby shipowners can get a real-time overview of fuel consumption and deviation detection for their ships“, Lange continues.

The goal of the collaboration is to create added value for customers so that, for example, the downtime of their mechanical components is reduced.

“This type of technology can be used in all industries, not just in the maritime sector and the only component we require for this is access to sensor data“, concludes Listou Ellefsen.

About Dipai

  • Dipai evolved from an innovation project under the auspices of the Department of Ocean Space Operations and Construction Engineering at Norwegian University of Science and Technology (NTNU) in Ålesund, to become a separate limited company.
  • The project has received financial backing from NTNU Discovery amounting to NOK 750,000 as well as NOK 500,000 from the Research Council of Norway.
  • Two of the three founders behind Dipai have a PhD from NTNU, in addition, several PhD students are also employed by the company on a part-time basis.
  • Dipai are experts in data analysis and their strongest competitive advantage is extensive expertise in artificial intelligence and machine learning.

Published: 20.11.2020