IntelliNord Mind


Machine Learning Driven Fault Diagnostics



What is IntelliNord Mind?

Life-Cycle Cost Analytics

  • Recognize repair debt.

  • Plan the device replacement schedule easily once the condition of devices is known.

  • Extend the life cycle of devices.

  • Detect device types that become too expensive to maintain.

Device Maintenance

  • Detect devices that are failing before it is too late.

  • Find out devices that have the same fault condition.

  • Reveal latent faults.

  • Improve the accuracy of field work.



Leading with knowledge!


AI Reseach with Helen/ Helen Electricity Network

Electricity meters tie up a considerable amount of capital and incur ongoing maintenance costs.

The study examined the suitability of artificial intelligence and machine learning for determining the condition of electricity meters.

The device’s health index can be used to make investment decisions. Another application is maintenance troubleshooting automation.

The devices detect faults in many ways and the data is collected in a reading system. The information to be utilized is e.g. energy time series, hourly data, device logs, self-diagnostics, power failure data, and component age data.

In the model developed in the study, data is continuously collected from the last ten days of the device. Data is also collected on device changes. This information acts as a human response because it is a human decision to change the device. The data is designed and converted to conform to machine learning algorithms.

The work compares the use of five different machine learning algorithms. A closer look ended with four algorithms: logistic regression, random forest, naive bays, and perceptron.

Of these, logistic regression, naive bayesia, and random forest are used to calculate the health index. Different models give very different probabilities for hardware failures, but they are parallel to each other. None of the models are overwhelmingly good.

The work developed is a fully automatic artificial intelligence system that is constantly learning. For example, the system can be scheduled to create a forecast model once a month. The system calculates the health index for each device on a daily basis and generates a health index report.

Read more about study below (document is in Finnish).

Helen Sähköverkko is one of the largest energy utility in Finland.

Kuntoindeksiraportti_public.pdf

Deploy the system in one day!


You will start to get results after 14 days learning period.