Real-time Process Monitoring for Froth Flotation Processes using Image Processing and Dynamic Fundamental Models

• Author / Creator

  Popli, Khushaal

• Froth flotation is a crucial and complex process, which has been used in various industries for the purpose of separation and beneficiation. Various classes of factors including chemistry, operational, or equipment influence the process outputs. Its complex nature makes it difficult to control and monitor the process. For instance, a small disturbance in any of these classes of factors propagates to the final process output by affecting various micro-scale sub-processes. In addition to this, process measurements are either accurate with slower sampling time, or less accurate and expensive for faster sampling times. The overall objective of this thesis is to develop a monitoring scheme aided with soft sensor models for online measurements for process outputs. Systematic study was undertaken in this thesis to develop monitoring schemes for mechanical cell flotations, starting from pure galena mineral flotation to synthetic mixtures of galena and quartz, and finally extending it to the real industrial Pb-Zn sulfide ore. A dynamic process model is required for the monitoring purposes. It was proposed that a fundamental model, providing the in-depth understanding of the process would perform better than empirical models to capture the disturbances in the process. For this purpose, a framework was developed for dynamic fundamental modeling that incorporates the mathematical relationships between micro-scale sub-processes and macro-scale transportation. All the significant sub-processes such as bubble-particle attachment, detachment, entrainment, and drainage, were included in the framework . Both pulp and froth phases were considered in the model development phase. Relationship between the froth visuals and process performance were explored to develop soft sensors for online measurements for the process outputs, which are, grade and recovery. A commercial package, VisioFroth, by Metso Minerals was used to extract real-time images and its features. Image features were used to develop soft sensor models for process outputs. Soft sensor model were developed based on machine learning algorithms, such as, Principal Component Regression, Partial Least Squares Regression, Random Forest, and Support Vector Machines. Soft sensor study was also extended to oil sands extraction process to demonstrate its applicability in other industries that use flotation on a regular basis. With bitumen being darker in color, it was more challenging problem as compared to the mineral flotation. However, set of robust soft sensor models, that were valid on various flotation conditions showed a promising potential in other real-time objectives such as model predictive control or real-time optimization. Finally, real time measurements were reconciled with the dynamic modeling framework for state and parameter estimation using extended Kalman filter. Estimation of model parameters that represent flotation sub-processes, provides real-time information about the process performance. Heuristics were developed for monitoring the process and identifying the disturbances though monitoring of parameter estimates. Various classes of disturbances were artificially created in batch flotation experiments in mechanical cell flotation. This included variation in feed particle size, reagent dosages, air flow rate, and impeller speed. Estimated parameters were successfully able to track the disturbances and identify its root for remedial actions. Developed scheme was also used to monitor the entrainment sub-process by decoupling the total recovery and identifying different components. Entrainment monitoring further helps in increasing the product grade while maintaining the desired recovery of the minerals. All the monitoring heuristics and soft sensor models were implemented and developed using batch flotation in a mechanical flotation cell.

• Subjects / Keywords

  • Fundamental modeling
  • Real-time monitoring
  • Support Vector Machines
  • Entrainment monitoring
  • Image Processing
  • Froth flotation
  • Extended Kalman Filter

• Graduation date

  Fall 2017

• Type of Item

  Thesis

• Degree

  Doctor of Philosophy

• DOI

  https://doi.org/10.7939/R3JH3DG68

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.