Online estimation of slag volume inside grate kiln pellet plant by stereo vision and machine learning
Purpose and goal
Slag is the largest preventer of continuous and efficient operation of grate-kiln pellets plants. Slag problems often cause shutdowns of the plant resulting in production losses and repairs. The restarting of the cold plant requires a significant amount of fuel for reaching production temperatures again. The project aims to develop a sensor based on stereo triangulation and machine vision for estimating the amount of slag inside the grate kiln pellet plant. The sensor will be used to monitor and minimize the slag formation by detecting its precursors for growth.
Expected results and effects
The expected effects of this project is increased understanding of how slag can be monitored, the slag formation mechanism, and how the slag growth can be minimized. The acquired knowledge from the sensor measurements is believed to result in increased plant efficiency, less shutdowns and less production losses at the pellet plant.
Planned approach and implementation
Within the initial phase of the project a market overview will be done to evaluate the available camera systems for stereo triangulation that can be adapted and applied for hot and dusty environments. From the acquired images and their depth information, an AI-model will be trained to detect the slag and together with the depth information its volume will be estimated. The developed prototype will be used and evaluated for the ability to prevent shutdowns due to the slag and minimize its buildup by newly acquired knowledge about the slag formation.