Magnetic Density Separation (MDS) is a novel method for the separation of shredded complex wastes [7, 8]. It uses a magnetically induced density gradient in a ferrofluid, which allows particles with different densities to follow different trajectories. Small variations in feeding direction, spread, and velocity can significantly affect the quality and purity of the separation. The MDS-based machine (MDX) studied in this research was developed by the Technical University of Delft and Myne Circular Metals B.V. After research on magnet configurations [9], fluid stabilization [10], fluid dynamics [11, 12], and the particle sliding phenomenon [8]. This thesis investigates how feeder geometry, vibration parameters, and material properties influence feeding performance and how this affects the separation. A Discrete Element Method (DEM) model was developed to simulate the behaviour of shredded copper cable feed as it is fed to the MDX by a vibratory feeder. The feed was modelled as a mixture of copper, aluminium, PVC, and PP/PE, represented by superquadric particles. Material properties and interaction properties were calibrated using literature and a Bulk Calibration Approach (BCA). A Design of Experiment (DoE), including a Definitive Screening Design (DSD), was used to calibrate sliding friction and restitution coefficients for all interaction parameters against three Key Performance Indicators (KPIs): discharge time, heap height, and angle of repose. The best-performing parameter set used a sliding friction of 0.8 and a restitution of 0.1, uniformly applied across all material types. The DEM simulations were validated against experimental tests, showing good agreement in mass flow but some deviation in feeding profiles. This calibrated DEM model was then used to test 20 feeder design variations across six base concepts, including straight and arced ends, varied feeder angles, the addition of particle-guiding walls, and different feeding bed interaction parameters. Feeding performance was quantified using four output parameters: horizontal and vertical insertion velocity, horizontal spread (modelled as a Gumbel distribution), and mass throughput. These were selected based on their known influence on separation purity in MDX. The results showed that higher vibration frequencies increased both feeding speed and particle spread. Arced feeder ends tended to increase forward particle velocity and spread, especially at larger radii, while the straight-ended design minimised these effects. The addition of a guiding wall helped reduce horizontal spread without sacrificing throughput. Ultimately, the best concept uses a straight end with a guiding wall to achieve low particle speeds and a narrow horizontal spread, with a throughput of 0.5 kg/s.

To improve the accuracy of Discrete Element Method (DEM) simulations for blast furnace materials, we present a structured calibration methodology for pellet and sinter particles based on nine key performance indicators (KPIs). For each material, the following process is followed: a Plackett-Burman sensitivity analysis to identify influential parameters, a Central Composite Design to develop polynomial regression models, and a comparison of multi-objective optimization techniques, including local optimization, genetic algorithms, and particle swarm optimization (PSO). This approach enables parameter input estimation of the DEM model. Experimental data is used to find input parameter values for the calibrated models, showing that PSO achieves faster convergence than GA and delivers accurate predictions at the original hopper height. However, discrepancies in the validation model suggest the need for refined parameter estimation to ensure robustness under varied conditions.

The escalating global demand for poultry products, coupled with the challenges of securing personnel for the often arduous tasks in poultry processing plants, has underscored the urgent need for automation in this sector. The inherent variability and delicate nature of poultry products, like chicken fillets, pose significant hurdles for traditional automation solutions. This research seeks to address this challenge by employing simulation-aided design to develop an automated solution for singulating and orienting chicken fillets. The study investigates existing technologies in poultry and other food industries and develops a practice-oriented Discrete Element Method (DEM) model to simulate chicken fillet behaviour. The model's efficacy is validated through a series of tests, including damping, bending, sliding, sensitivity analyses and real-life experiments. The insights gained from these simulations inform the design of an automated solution, which is then iteratively refined and visualised in the DEM software. The proposed solution successfully demonstrates the potential of simulation-assisted design to automate complex tasks within the poultry processing industry, paving the way for increased efficiency, productivity, and reduced reliance on manual labour. The research also underscores the importance of considering the unique characteristics of poultry products in developing effective automation solutions, highlighting the need for further exploration into advanced modelling techniques and material and interfacing equipment behaviour analysis.