A framework to offer high value manufacturing through self-reconfigurable manufacturing systems

The High Value Manufacturing (HVM) sector is vital for developed countries due to the creation of innovative products with advanced technology that cannot be reproduced at the same cost and time with traditional technology. The main challenge for HVM is to rapidly increase production volume from one-off products to low production volume. This requires highly flexible manufacturing systems that can produce new products at variable production volumes. Current manufacturing systems, classified as dedicated, flexible and reconfigurable systems, are limited to produce one type of product(s), within a production volume range and have fixed layouts of machines. Thus, there is a need for highly flexible systems that can rapidly adjust their production volume according to the production demand (i.e. main HVM challenge). Therefore, a novel manufacturing framework, called INTelligent REconfiguration for a raPID production change (INTREPID), is presented in this thesis. INTREPID consists of a user interface and communications platform, a job allocation system, a globally distributed network of Reconfigurable Manufacturing Centres (RMCs), consisting of interconnected factories, and Self-Reconfigurable Manufacturing Systems (S-RMSs). The highly flexible S-RMS consists of movable machines and Mobile Manufacturing Robots (MMRs). The novelty of the S-RMS is its capability of forming layouts bespoke to the current production needs. The vision of INTREPID is to offer global HVM services through the network of RMCs. The job allocation system determines the best possible RMCs or factories to perform a job by considering the complexity of the production requirements and the status of the available S-RMSs at each factory. The planning of the production with S-RMS is challenging due to its high flexibility. The main example of this flexibility is the possibility to create layouts bespoke to current production needs. Yet, this flexibility involves the challenges of determining allocations and schedules of tasks to robots and machines, positions to manufacture, and routes to reach those positions. In manufacturing systems with fixed layouts, production plans are determined by solving a sequence of problems. However, for the S-RMS, it is proposed to determine production plans with a single problem that covers the scheduling, machine layout and vehicle routing problems simultaneously. This novel problem is called the Scheduling, positions Assigning and Routing problem (SAR) problem. In order to determine the best possible production plan(s) for the S-RMS, it is necessary to use optimisation methods. Dozens of elements, characteristics and assumptions from the constituent problems might be included in the formulation of the SAR problem. Elements, characteristics and assumptions can be considered as decision variables on whether to include or not the elements and characteristics and under which assumptions in the formulation. There are two types of decision variables. Fundamental variables are natural to the SAR problem (e.g. manufacturing resources, factory design and operation), whilst auxiliary variables arise from the aim to simplify the formulation of the optimisation problem (i.e. time formulated as discrete or continuous). Due to the large number of decision variables, there might be millions of possible ways to formulate the SAR problem (i.e. the SAR problem space). Some of these variants are intractable to be solved with optimisation methods. Hence, before formulating the SAR problem, it is necessary to select a problem(s) that is realistic to industrial scenarios but solvable with optimisation methods. Existing selection methods work with pairwise comparisons of alternatives. However, for a space of millions of SAR problems, pairwise comparisons are intractable. Hence, in this thesis, a novel Decision Making Methodology (DMM) based on the controlled convergence method is presented. The DMM helps down-selecting one or a few SAR problems from millions of possible SAR problems. The DMM is demonstrated with a case study of the SAR problem and the results show a significant reduction of the reviewed SAR problems and the time to select them.