METROSCOPE patented and awarded diagnosis comes from years of research at EDF LAB, one of the biggest industrial R&D centers in Europe.

It allows us to take the best from the physical equations carried by the digital twin to explain the measurements on the process. Not the other way around!

In comparison with Big Data solution that looks for valuable information in a profuse environment, METROSCOPE is facing a small but structured data problematic.

Thus, METROSCOPE is respecting the natural way of expert reasoning using an established knowledge of the physics and design of the plant.

Markov chains


METROSCOPE AI relies on Markov chains, a mathematical approach that allows METROSCOPE to take into account the uncertainties of the measurements. It is the key to a reliable diagnosis.

A Markov chain is a stochastic model describing a sequence of possible events where the probability of each event depends only on the state attained in the previous event.

Saying that, what have we said? METROSCOPE will explore the different scenarios in an iterative way weighted by probabilities. The Markov chain will automatically guide the exploration of the scenario to the “hot spots”, e.g. the most likely diagnosis.
METROSCOPE engine is powered by Markov Chain, just like Google page rank!

In the video hereunder, the 5 percentages represent the probabilities of 5 different industrial hazards and each of the 200 points follows a Markov Chain.

When more than 95% of the points gravitate around the same hazard, METROSCOPE will raise an alert and show the diagnosis.

You do not need to get the Maths to see the beauty of it!

Sensitivity and Specificity

2 indicators are key to understand the performance of a diagnosis test such as METROSCOPE: the sensibility and the sensitivity.

The Sensitivity (also called the true positive rate) measures the proportion of actual positive diagnosis that are correctly identified as such (e.g. the percentage of hazards correctly identified).

The Specificity (also called the true negative rate) measures the proportion of actual negative diagnosis that are correctly identified as such (e.g. the percentage of hazards not happening but correctly identified as so).

A high sensitivity will result in better detection capacities but will also lower the specificity and generate false alarms!

On the contrary, a high specificity will ensure the reliability of the diagnosis.


At METROSCOPE we have chosen to maintain a high specificity of our diagnosis, it means that METROSCOPE is, by design, very reliable.

Which is a must have for any industrial operator.

From our experience, we have less than 10 % of actual false alarms on-site with METROSCOPE and we still have a detection that is up to 5 times faster than other approaches.

You can entrust METROSCOPE!

Algorithm reinforcement

Will the diagnosis performance improve in time? Can we detect new hazards and interpret new measurement? By design METROSCOPE diagnosis is made to evolve in time.

It is easy to add new measurement or new hazards to the scope of knowledge of METROSCOPE. Hence, in time, METROSCOPE will be able to always detect more scenarios, with a growing performance.