min read

Detecting silent failures in the making for our Heavy Industry customers

Our main product, Canopy, is built on top of very advanced AI models that are agnostic to the application. This means that they can be used to reduce machine downtime in various industries.

Our main product, Canopy, is built on top of very advanced AI models that are agnostic to the application. This means that they can be used to reduce machine downtime in various industries such as wind power where we have used them to detect failures of internal components ahead of time as well as power performance issues (read more about here).

Our normality models due to the way they are trained can detect any types of faults, i.e. they don’t require a labelled dataset of annotation faults. This means that we can easily detect faults that have never happened before!

In this blog, we will be presenting two critical detections made by our Normality models (more information here) and delivered timely to our industrial customers via Canopy. The two detections present two important cases that are related to clogged filters.

Clogged filters are silent problems that can easily get out of hand and lead to catastrophic failures and consequently, to factory downtime.

Detection case: example 1

A factory is already in itself a mind-boggling complex mesh of circuits, pipes and machinery. One definitely requires the right tooling to not get lost in the jungle! Thanks to our AI models that always keep a tab on the various sensors of the factory, our users will be alerted as soon as something starts behaving abnormally!

Figure 1 - A silent killer waiting for the perfect time to cause factory downtime.

The filter from the image above caused a temperature to quickly increase reaching values 50% higher than expected. This was all timely informed to our users via Canopy.

Figure 2 - Temperature increase detection in Canopy.

Our users were alerted to the fact that this steadily growing deviation was happening! Due to our clear alarms, they were able to easily identify the right course of action to stabilise the abnormally hot temperature in their next planned maintenance window: cleaning of the filter (indicated by the blue arrow 🔽 in Figure 1). This simple action completely fixed the ongoing issue and brought the temperature to its expected values during the planned corrective action on the 27th (see in Figure 2).

Figure 3 - Debris was removed from the clogged filter that was causing a dangerous temperature increase.

Since the maintenance team of the factory was alerted early they successfully avoided unplanned downtime and potentially higher production losses due to a catastrophic failure of the equipment.

Detection case: example 2

Another detection made by our AI models was in the cooling system of a massive cluster of IGBT-based power converters (shown below). These converters are of paramount importance to run a non-stop production line. If they fail the entire factory line will be shut down. Factory processes where each hour is worth thousands of euros, each minute is extremely valuable!

Figure 4 - A small portion of the large cabinet filled with dozens of power converters.

Canopy alerted our users that the temperatures of all power converters were above their expected values (see below an example for two power converters). Furthermore, the deviations were also increasing over time meaning that our users had to act quickly to bring the growing anomaly to a halt.

Figure 5 - The power converter temperatures were above what our AI models predicted.

After analysing the deviations that our AI models provide at the sensor level and our alarms, the maintenance team understood that the problem had to be located at a place that could impact the temperature of all converters. This line of thought took the maintenance team to investigate the presence of dirt in the cooling system filters.

Once again the maintenance team was able to easily pinpoint the location of the problem! After cleaning the filter (shown below) the temperatures of all power converters were brought back to their expected values (see in the above image around Wednesday 26th). Another detection that prevented unplanned downtime!

Figure 6 - Dirt located inside the filter led to the continuously growing abnormal temperatures of the power converters.

In future posts, we will demonstrate how our AI-based product is helping factory maintenance teams to be alerted for motor bearings and couplings that require new grease pumps and how we help an industrial customer keep the chemical concentrations at the right level to increase product quality. Stay tuned for more!

Silvio Rodrigues

CIO & Co-founder

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Suspendisse varius enim in eros elementum tristique. Duis cursus, mi quis viverra ornare, eros dolor interdum nulla, ut commodo diam libero vitae erat. Aenean faucibus nibh et justo cursus id rutrum lorem imperdiet. Nunc ut sem vitae risus tristique posuere.

Back to blogs