News
Carrying out periodic inspections in industrial processes is of great importance for the success of an operation. It is through these that the detection of anomalies and other failures becomes possible before they turn into bigger problems. With this in mind, Pix Force joined forces with the company Eneva for a project in an important research and development program of ANEEL (National Electric Energy Agency). In this text, we present the details of a Pix Force success story. Enjoy your reading!
Speak with one of our specialists and discover how Pix Force can transform your business
What was the project and its objective?
The project's main objective was to develop an automated, applicable, and scalable product for detecting anomalies in pipe conveyor belts. The goal was to perform preventive maintenance to minimize negative impacts during use, prioritizing the safety of everyone involved in the process. In addition, the project anticipated a reduction in costs caused by spontaneous interruptions. For the development of the research, a conveyor belt approximately 4 kilometers long and with more than 24,000 rollers was selected. This belt runs from the port of Itaqui to the Eneva Thermoelectric Power Plant, both located in the municipality of São Luís, in Maranhão. During the research, some problems were noted in the process. Among them, we list: • Extreme heating in the rollers
• Belt jamming
• Excessive friction with the belt
• Cuts and, quite often, fires
• Rotation of the overlap
All of these end up causing larger problems, including the total stoppage of the belt. This, in itself, results in higher maintenance and correction costs.
And how can these problems be solved?
To inspect the aforementioned problems, visual and auditory observations are made during the conveyor's operation. From there, data such as the overlap position or a stopped roller are found. These processes, however, are carried out by a field team that needs to cover the path of interest. The arduous path, which goes through terrain changes, ends up putting the worker's safety at risk, and therefore, an optimization of the entire system was proposed. The optimization of the process was achieved through a set of hardware and software on a drone equipped with a camera and sensors. The equipment was able to obtain information from the electromagnetic spectrum in the visible (RGB) and thermal infrared wavelengths. In addition, an image recognition and classification system was used to analyze and point out potential failures. Thus, the R&D was segmented into four stages: • Development of an application capable of automating the drone's flight
• Development of machine learning algorithms capable of identifying overheated rollers
• Development of machine learning algorithms capable of identifying the position of the overlap on the belt
• Development of a web platform to visualize the generated results
A topographic study of the area of interest was also conducted to obtain a 3D model of the belt and, thus, automate the drone and design more precise flight routes. This process was carried out with three flight routes and three takeoff points along 4 km. • Read more: "IoT in industry: See success stories that will inspire you"
The application
The technologies developed for the application were made on the Android platform with a Kotlin/Java programming language. The purpose of the app was to have the functionalities and information of the drone during flight, having as options to execute routes or cancel the flight in case of an unforeseen event. Additionally, functionalities for downloading data into the processing and analysis system were added to the application.
The algorithms
The development of the computer vision algorithms started from convolutional neural network models. This presents greater assertiveness in detection and reduces the number of constants in the image projection. Another model involves area or structural detection, facilitating the detection of missing rollers. Thus, it becomes possible to work with an estimated model of rollers present in the region of interest. For this project, we used the RedHat Keycloak platform, which offers the option to support multiple applications and sessions with different access types.
Project Conclusions
With this project, it became possible to identify problematic rollers in an agile and safe manner. The initial objectives were successfully met and managed to show a scalable solution for various sectors. Furthermore, the project included a platform to optimize flights, promoting greater accuracy in the results. Pix Force is a company that seeks innovation in various processes, and you can learn more about our work by visiting our website.
Fabio Caraça
Fábio Caraça is the Chief Growth Officer at Pix Force. He leads Pix Force's transformation into a scalable SaaS operation, combining strategic vision, culture, and high-impact execution.
