Crisis averted: How our AI-powered services helped prevent a factory fire

This audio was created using Microsoft Azure Speech Services

Maintenance

Schneider Electric’s predictive modelling provides an early warning of critical risks, helping businesses manage their electrical assets more efficiently.

Behind the brands we encounter every day is a massive infrastructure – factories, offices, warehouses, and transport networks – that makes it possible to provide their products and services. Even further from view are the energy systems that power this infrastructure, which increasingly rely on electricity. Out of sight doesn’t mean unimportant, however – and recent remedial action made possible by Schneider Electric’s AI-powered analytics shows just how much is at stake.

One of our customers, a global food and drink manufacturer, runs a large factory in Latin America. The facility runs 24 hours a day, seven days a week, employing 4,000 workers and producing more than a million dollars’ worth of produce a day. Any failure in the power supply has a big impact. This was underlined in 2019, when a medium voltage cubicle exploded and caused the factory to shut down. Closing for just a day and half to replace the component meant the company took a significant hit to its turnover.

AI-powered analytics

For this company, maintaining stable production levels is crucial. So, shortly after the fire, it decided to begin monitoring its electrical systems more effectively using our digitally enabled service plans. This involved installing thermal sensors – both on the components of the power system, and in the surrounding environment. The IoT-connected sensors measure the temperature at key locations to provide continuous readings, accessible from a cloud-based dashboard. These can then be compared against the levels typically seen for similar connections and electrical loads to highlight possible concerns.

This isn’t always a simple matter because the temperatures can vary widely over time. To enable greater accuracy, the analytics also incorporate a second dimension, powered by AI. A machine learning model is used to compare patterns in the expected and actual temperatures, helping to identify any risk at a very early stage.

As soon as the monitoring was set up at this site, it became crystal clear that there was a problem at one transformer, where the sensors were giving very high readings. The temperature was reaching more than 125°C (over 257°F), when it should have been around 70°C (158°F) and the safe limit, in theory, was 90°C (194°F). The company’s engineers inspected the component as soon as possible, supervised remotely by our experts. They found that a screw at one connection had been incorrectly tightened, with the result that an important cable was connected too loosely. This made it harder for the current to pass through, which in turn was causing the temperature at the connection to rise.

Maintenance

Detection of an overheating on a Dry Transformer

Expert advice

This created a dangerous situation. When copper electrical parts heat up, it can make it even harder for current to flow through them – which can then lead to even higher temperatures, in an ever-increasing spiral. In this case, the cable was already so hot that the insulation surrounding it was melting – and that was with the transformer only loaded at 50 per cent capacity. Any increase in the electrical load could easily have caused an explosion.

On top of this came another unwelcome finding. As the fault was being investigated, another serious – and separate – problem was identified at the same transformer. While advising the customer remotely on the phone, a Schneider Electric expert identified signs of partial discharge. This problem arises when cables are placed too closely to each other. The exchange of electrical charge between the cables gradually erodes the insulation and can, again, potentially cause a major explosion. Fortunately, with the problem spotted in good time, the customer was able to change the cables and re-install them a safe distance apart.

An early warning system

The company’s investment paid off immediately. The warning offered by Schneider Electric’s predictive analytics, together with our expert guidance, was crucial in helping avoid not one, but two further costly shutdowns. And while some activity at the site could well have resumed within a day or so, replacing a broken transformer would likely have meant many months of running the system at dangerously high-risk levels. Having seen the benefits, the customer is now in the process of extending its monitoring – installing additional sensors to provide richer data.

AI Agent, Maintenance

Early warning system from Schneider Electric’s predictive analytics

In the way they work, our digitally enabled service plans such as EcoCare*, can be compared to a wearable fitness device, like a smart watch. By tracking inputs including heart rate and movement, these devices generate accurate models about things like what sport someone is playing, whether they are awake or asleep, and their breathing rate.

This information can also be further analyzed to provide a range of indexes and recommendations – such as suggesting whether activity or rest would be best on any given day. In this way, wearable trackers can act as an early warning system to help us manage and reduce risks. They might indicate where changes in lifestyle or diet could contribute to maintaining good health, for example – reducing the need for medical intervention in future or focusing it where it is most required.

What fitness trackers do for the human body; Schneider Electric’s AI-powered analytics provide for power systems around the world. It can be difficult to physically examine electrical equipment without shutting down large parts of the system. But our real-time data and accurate modelling gives engineers the information they need to keep everything running smoothly – making timely adjustments to maintain performance while avoiding disruptive emergencies. If a component at this factory begins to overheat in future, for example, the engineers will know about it at much earlier point.

Of course, human input remains essential – as we can see in this case, where contributions from our remote support team were crucial in helping the business detect, confirm, and resolve the issues that emerged. But equally, their guidance was made possible because of the insights from the analytics. In this way, the steps taken to shore up the factory’s power supply encapsulate our wider approach: experts and technology working together to support better, more reliable services.

*Please verify the availability of EcoCare in your region through a local services sales’ representative. If EcoCare is not yet available, you can start leveraging EcoStruxure Service Plan.

Tags: , , , , ,

Add a comment

All fields are required.

This site uses Akismet to reduce spam. Learn how your comment data is processed.