Abstract:
To do away with the hazardous, tedious, repetitive, and export-dependent gold panning
procedure currently practiced by artisanal and small-scale miners in developing countries
an automated system for the gold panning procedure is presented in this thesis. Gold panning is one of the gravimetric separation processes which separate particles of greater specific gravity (particularly gold) from mineral waste. This separation method is prevalently used in low-tech areas and mercury is incorporated to make for effective and efficient gold recovery. The mercury used exposes the miners to hazardous health conditions and increases mercury pollutions. Little technical effort and knowledge are available to provide solutions adaptable to the African context. This study is devoted to the technical optimization of the gold washing process and hand-picking process involved in the current panning procedure through mechanization and automation. The design was achieved through the use of the V model for mechatronic system design (VDI 2206 guidelines). Morphological analysis was used to adapt and make use of the already existing concepts and designs to mechanize and automate the manual, expert-dependent, and tedious gold panning process. The design adopted was based on the usage of visual data to control the separation and picking process.
In the automation of the gold panning process and robotic handling, it is important to
identify and locate the gold particles in the images captured by the image sensor. Color
image thresholding in the CIELAB color space was developed as a front-end technique to
automate the visual feature identification and localization of gold particles in the pan. 3D
CAD models of the designed system were developed in Solid works. Finite element analysis of the designed washer concept was performed in ANSYS 2019 to precisely calculate material stress, strain, deformation, contact, and safety factor. This was done to predict the behaviour of the components and assembly under given boundary conditions and loading in a quasi-real scenario. Kinematic modeling of the handling system was performed using the Denavit Hartenberg convention and analytical methods. Physical modeling using MATLAB, Simulink, and SimScape was done to perform dynamic modeling and simulations of the handling system. Stateflow charts were developed to model state machines and flowcharts which were used in task planning to interface various components used in the vision-based control and picking process. Simulation and modeling results showed that the design is feasible, achievable, and efficient.
