Modeling and optimization of a multi-stage shredding-separation systems for recycling

This Thesis is the result of the work performed in collaboration with the Massachusetts Institute of Technology. Its aim was the realization of an Industrial Engineering approach for recycling plants. Justification for it can be found on the one hand in the dramatic development of recycling market, due to the increasing in products' life and in more and more strict legislation, and on the other in the lack of tools and studies on the topic. In recycling facilities, end-of-life products are input to the plant, where they first undergo the process of comminution, in which they are cut into pieces with the purpose of obtaining particles made as much as possible of one material. After this process, particles are separated thanks to their physical properties, with the purpose of collecting the materials contained in them. In these systems the connection between the process level and the system level is very strict, as the output of each process strongly influences the choices at the next stations and their output. Our work can be divided into two parts: a process level and a system and optimization level. The former includes a modeling effort of two processes: the comminution operation, using results of previous work especially in mineral processing, and the eddy current separation, for the collection of non-metals. At the system level, a model was realized to characterize the performance of a recycling system using a transformation matrices approach keeping a very high level of detail on the process. Analytical solution for a special case was developed, and solution algorithms were realized for more complex situations, including systems with rework flows of materials. An optimization method based on simulated annealing was developed and tested, with the aim of choosing the best parameters for all the machines in the system in a global optimization approach. The innovative contribution of this work can be found in the methodology that connects processes and system in a way that was not tried in the field before but is required by the features of these systems, and in the optimization method that jointly tunes the parameters of all the machines in the system.