Time and Date: 10:15 - 11:55 on 3rd June 2015
Chair: Timoleon Kipouros
|377|| A Solution for a Real-time Stochastic Capacitated Vehicle Routing Problem with Time Windows [abstract]
Abstract: Real-time distribution planning presents major difficulties when applied to large problems. Commonly, this planning is associated to the capacitated vehicle routing problem with time windows (CVRPTW), deeply studied in the literature. In this paper we propose an optimization system developed to be integrated with an existing Enterprise Resource Planning (ERP) without causing major disruption to the current distribution process of a company. The proposed system includes: a route optimization module, a module implementing the communications within and to the outside of the system, a non-relational database to provide local storage of information relevant to the optimization procedure, and a cartographic subsystem. The proposed architecture is able to deal with dynamic problems included in the specification of the project, namely: arrival of new orders while already optimizing as well as locking and closing of routes by the system administrator. A back-office graphical interface was also implemented and some results are presented.
|Pedro Cardoso, Gabriela Schütz, Andriy Mazayev, Emanuel Ey, Tiago Corrêa|
|25|| Fast Multi-Objective Optimisation of a Micro-Fluidic Device by using Graphics Accelerators [abstract]
Abstract: The development of technology that uses widely available and inexpensive hardware for real-world cases is presented in this work. This is part of a long-term approach to minimise the impact of aviation on the environment and aims to enable the users both from industrial and academic background to design more optimal mixing devices. Here, a Multi-Objective Tabu Search is combined with a flow solver based on the Lattice Boltzmann Method (LBM) so as to optimise and simulate the shape and the flow of a micro-reactor, respectively. Several geometrical arrangements of a micro-reactor are proposed so as to increase the mixing capability of the device while minimising the pressure losses and to investigate related flow features. The computational engineering design process is accelerated by harnessing the high computational power of Graphic Processor Units (GPUs). The ultimate aim is to effectively harvest and harness computing cycles while performing design optimisation studies that can deliver higher quality designs of improved performance within shorter time intervals.
|Christos Tsotskas, Timoleon Kipouros, Mark Savill|
|294|| Multi-objective Optimisation of Marine Propellers [abstract]
Abstract: Real world problems have usually multiple objectives. These objective functions are often in conflict, making them highly challenging in terms of determining optimal solutions and analysing solutions obtained. In this work Multi-objective Particle Swarm Optimisation (MOPSO) is employed to optimise the shape of marine propellers for the first time. The two objectives identified are maximising efficiency and minimising cavitation. Several experiments are undertaken to observe and analyse the impacts of structural parameters (shape and number of blades) and operating conditions (RPM) on both objective. The paper also investigates the negative effects of uncertainties in parameters and operating conditions on efficiency and cavitation. Firstly, the results showed that MOPSO is able to find a very accurate and uniformly distributed approximation of the true Pareto optimal front. The analysis of the results also shows that a propeller with 5 or 6 blades operating between 180 and 190 RPM results in the best trade-offs for efficiency and cavitation. Secondly, the simulation results show the significant negative impacts of uncertainties on both objectives.
|Seyedali Mirjalili, Andrew Lewis, Seyed Ali Mohammad Mirjalili|
|502|| Distributing Fibre Boards: A Practical Application of Heterogeneous Fleet Vehicle Routing Problem with Time Windows [abstract]
Abstract: The Heterogeneous Fleet Capacitated Vehicle Routing Problem with Time Windows and Three-Dimensional Loading Constraints (3L-HFCVRPTW) combines the aspects of 3D loading, heterogeneous transport with capacity constraints and time windows for deliveries. It is the first formulation that comprises all these aspects and takes its inspiration from a practical problem of distributing daily fibre board deliveries faced by our industry partner. Given the shape of the goods to transport, the delivery vehicles are customised and their loading constraints take a specialised form. This study introduces the problem and its constraints as well as a specialised procedure for loading the boards. The loading module can be called during or after the route optimisation. In this initial work, we apply simple local search procedures to the routing problem to two data sets obtained from our industry partner and subsequently employ the loading module to place the deliveries on the vehicles. Simulated Annealing outperforms Iterated Local Search, suggesting that the routing problem is multimodal, and operators that shift deliveries between routes appear most beneficial.
|S Pace, A Turky, I. Moser, A Aleti|
|679|| Performance Comparison of Evolutionary Algorithms for Airfoil Design [abstract]
Abstract: Different evolutionary algorithms, by their very nature, will have different search trajectory characteristics. Understanding these particularly for real world problems gives researchers and practitioners valuable insights into potential problem domains for the various algorithms, as well as an understanding for potential hybridisation. In this study, we examine three evolutionary techniques, namely, multi-objective particle swarm optimisation, extremal optimisation and tabu search. A problem that is to design optimal cross sectional areas of airfoils that maximise lift and minimise drag, is used. The comparison analyses actual parameter values, rather than just objective function values and computational costs. It reveals that the three algorithms favoured various extents of explorations on the different parameters.
|Marcus Randall, Tim Rawlins, Andrew Lewis, Timos Kipouros|