Artificial Intelligence applied for anomalies detection – Step by step 2
Experiment #207 – Discover the Artificial Intelligence applied for anomalies detection
Step by step 2
We have performed three analyses, one for each case of use:
On this panel we can see the state of the solar panels of our solar farm in reference to its state of cleanliness and, therefore, the amount of energy that it will provide us with.
There are three states: Complete when the panel is clean, Moderate when it has a percentage of less than 66% with dirt and Poor when the dirt exceeds 66%.
On the general map, we can see at a glance the state that predominates in our plates, as well as identify possible locations where more dirt accumulates for various reasons (proximity to roads, mountainous areas, etc).
Also, we can see a pie chart in which we can evaluate the number of solar panels with respect to their condition.
On the other hand, we have two metrics:
We want to evaluate if the number of dirty honeycombs (Moderate + Poor) is higher than the number of clean solar panels (Full). If it is, we should launch a cleaning event as soon as possible.
Finally, the number of Poor solar panels should always be less than 20% of the total number.
In this panel we can perform an analysis to trace the use/non-use of PPE in a construction when the use of PPE is mandatory.
We can apply filters such as date or Person.
We can see two pie charts:
– The first one shows the absence of PPE by zone. We can clearly see how in zone 2 there is a higher probability of not wearing PPE than in zone 1. We could have as many zones as we wish depending on the placement of the security cameras.
– On the other hand, the second foot chart shows the number of people wearing PPE versus those seen without PPE.
Likewise, the bottom bar chart shows us at a glance the use/non-use of PPE by date. We can clearly see that the hottest months of the summer (July and August) are when PPE absences start to be found.
In this panel, we can perform an analysis to identify possible anomalies such as rust in windmills.
We have 4 states: No rust when there is no rust present, Blades rust if the rust is on the blades, Shaft rust if it is on the shaft and “both shaft and blades rust” in case the rust is in both places.
In the graph on the left we can see at a quick glance the condition of our windmills in relation to the presence of rust with the 4 states mentioned above.
The bottom chart shows a summary of the four states versus the number of windmills.
Finally, we have two interesting metrics:
– The first evaluates whether the number of rusted windmills (rust blades + rust shaft + both shaft and blades rust) is higher than those without rust. In this case, we see that this number is much higher, so we should take action as soon as possible.
– Finally, the second metric tells us if those “dangerous” mills, i.e. those with rust on the shaft (fust rust + both blades and shaft rust) is higher than 20% of the total.
The enormous number of applications of this project is undeniable. In addition to the three use cases already mentioned, we can also use it for, for example:
– Identify breaks in systems
– Identifying dampness
– Identification of anomalies in assembly lines
– Identifying fluid leaks in tanker fleets