The maintenance robot

Cognitive robots and artificial intelligence will change the way we do maintenance. Cognition means thought, ability to learn, acknowledge and draw conclusions. Why would a robot with these capabilities influence the way we do industrial maintenance?

To understand this we need to consider all the different maintenance philosophies that have existed since the industrial revolution started in 1750. It all began with a breakdown-based maintenance philosophy. This philosophy was based on not carrying out maintenance at all, but using the equipment until it broke down. Expensive repairs and unstable facilities were the result.

Then a time-based philosophy was developed. It was based on regular activities being performed on a regular basis. This leads to spending a lot of money on unnecessarily maintaining equipment that is in order, while faults that develop between the fixed intervals lead to breakdowns.

Resource use and breakdowns led to the creation of a condition-based maintenance philosophy. This is a philosophy in which you use sensors to monitor the equipment and only carry out maintenance when the condition indicates that it is necessary. Condition monitoring reveals the problem, but not necessarily its root cause.

Then a ‘predictive’ maintenance philosophy saw the light of day. It involved being one step ahead of the problems by analysing large quantities of data in order to predict faults, and at the same time find the root cause of recurring problems. This philosophy has, until now, had the disadvantage of being too costly and unpredictable to be used successfully in large plants. Despite considerable resources spent on analyses, the equipment has continued to break down without us necessarily understanding why.

One of the reasons for this is that 80 % of all available data in the world is unsorted. This means that men must manually collect, sort and edit the data in order to analyse it. Our brains have natural limitations for processing large amounts of data, therefore we often reduce the data so much while editing it that the information we are looking for is sorted away in the process. For that reason, a robot solution with artificial intelligence and cognitive abilities would make a great difference. The robot is capable of interpreting unsorted data in different formats and identifying patterns that have previously remained hidden to men.

A cognitive robot solution starts at nothing, and in that sense it is comparable to a newborn baby. Absolutely everything must be learned. The learning algorithms and access to data determine how smart the robot solution will be. The more data you can feed the robot, the more intelligent it will become. At Karsten Moholt AS we are currently working on developing a prototype of a cognitive maintenance robot. We will send a robot to study maintenance through our historical data accumulated over 70 years.

These are unsorted data in the form of service reports, root cause analyses, condition data and condition analyses, in different formats.