Video Analytics has been used since 2018, mainly for industrial safety tasks. The original idea was to find a way of monitoring compliance with the Ten Rules that Save Lives and now, the technical team has found fresh opportunities for its application.
The system first used to analyze possible safety deviations using video, which contributed to ensuring social distancing in times of COVID-19, has incorporated a new application. This is one more step in the path towards a digital transformation that the company is taking.
A new Video Analytics algorithm is now available to detect fires and add an additional containment barrier to the system currently in operation throughout the plants. So far, it has been implemented in Ternium Argentina’s San Nicolas and Ensenada plants as well as in Ternium Mexico’s Guerrero mill, and the objective is to replicate it throughout the company's other facilities.
Fire is one of the main risk factors for Severity 4 incidents. The algorithm offers an additional containment barrier to the current fire detection system by adding more detectable spaces, using the network of Video Analytics cameras. When the algorithm detects a fire, an audiovisual alert goes off in the fire control room, and the images are simultaneously projected on a screen for viewing. “This is a great complement to our fire detection systems, as they not only reach places that weren’t covered by the systems but also, by showing the emergency personnel team the live photo sequence they can immediately access the location with a clear vision of what is happening,” explains Marcos Fernández, Safety Engineer. These images make it possible to identify the type of fire and focus the response action. The objective is not to replace the current standard and certified fire detection systems, but to complement existing ones and take advantage of the coverage offered by the installed capacity of video cameras.
José Lacarra, Risk Prevention Supervisor, says that “this is a major technological advance for fire prevention functions. It means we can immediately identify a problem developing in a specific area in our plant, which may or may not have fire detection and extinguishing system. In addition, the strategic distribution of cameras provides us with a vision of the dimensions of the event and if there are people involved, this means that our participation can be more effective. It’s a dynamic, adaptive system which will surely offer opportunities for improvement over time.”
In 2020, Ternium's Video Analytics system was trained with algorithms developed in-company to monitor social distancing and spot whether people were using facemasks or not, in efforts to adapt the company’s industrial facilities and operations to work in the pandemic.
"The algorithms we developed in-house will allow early fire detection, contributing to the prevention of incidents and accidents, a key element in Ternium's Occupational Health and Safety policy," comments Jorgelina Wirsch, Industrial System Manager and Quality from Argentina.
Andrés Gómez, Ternium Product Development Manager, explains that: “We started training the algorithm in October 2020 using classified images (and videos) of real fires and fake fires (incandescent lighting, flashes, different lights, etc.) from Ternium's mills and third-party sources, today totaling more than 14,000 images. To do so, we used computer vision techniques and neural networks to detect fire possibilities in Ternium employing camera images based on shape, intensity, movement, as well as false and positive cases.”