Showing 6 results for Csébfalvi

A. Csébfalvi,

Volume 2, Issue 1 (3-2012)

Abstract

This paper provides a test method to make a fair comparison between different heuristics in structure optimization. When statistical methods are applied to the structural optimization (namely heuristics or meta-heuristics with several tunable parameters and starting seeds), the "one problem - one result" is extremely far from the fair comparison. From statistical point of view, the minimal requirement is a so-called "small-sample" according to the fundamental elements of the theory of the experimental design and evaluation and the protocol used in the drug development processes. The viability and efficiency of the proposed statistically correct methodology is demonstrated using the well-known ten-bar truss on a set of the heuristics from the brutal-force-search up to the most sophisticated hybrid approaches.

A. Csébfalvi,

Volume 2, Issue 2 (6-2012)

Abstract

The cumulative resource constraints of the resource-constrained project scheduling problem (RCPSP) do not treat the resource demands as geometric rectangles, that is, activities are not necessarily assigned to the same resource units over their processing times. In spite of this fact, most papers on resource-constrained project scheduling mainly in the motivation phase use a strip packing of rectangles (SPR) like visualization to illustrate the resource allocation. A novice researcher inspired by the "artistic" SPR visualization may think that the "rectangles" are essential elements of the RCPSP, and that the RCPSP is a special counter-intuitive strip packing problem (SPP) which can be solved without explicitly defined strip packing constraints. In this context "artistic" means, that we have to use a "drawing tool" to produce a SPR like visualization, because the standard model of the RCPSP knows nothing about the rectangles. In the RCPSP, the rectangles can be torn vertically and horizontally, which is absurd in the SPP, and the existence of a cumulative solution is only a necessary but not sufficient condition of the existence of the SPR like visualization, as proven by several researchers. Therefore the popular SPR visualization is theoretically wrong and misleading, and hides a real problem, which is connected to the dedicated resource assignment. In this paper, we prove that replacing the rectangles with a set of strips with unit height we can always generate a theoretically correct strip packing of strips (SPS) like dedicated assignment, where dedicated means that each demand unit is served by exactly one resource unit over its duration without "hidden" transfer time and cost.

A. Csébfalvi,

Volume 2, Issue 3 (7-2012)

Abstract

In this paper we present a unified (probabilistic/possibilistic) model for resource-constrained project scheduling problem (RCPSP) with uncertain activity durations and a concept of a heuristic approach connected to the theoretical model. It is shown that the uncertainty management can be built into any heuristic algorithm developed to solve RCPSP with deterministic activity durations. The essence and viability of our unified model are illustrated by fuzzy examples presented in the recent fuzzy RCPSP literature.

A. Csébfalvi , E. Szendrői ,

Volume 2, Issue 4 (10-2012)

Abstract

This paper presents an experimental investigation of the Sounds of Silence (SoS) harmony search metaheuristic for the multi-mode resource-constrained project scheduling problem (MRCPSP) using a pre-optimized starting repertoire. The presented algorithm is based on the time oriented version of the SoS harmony search metaheuristic developed by Csébfalvi et al. [1] for the single-mode resource-constrained project scheduling problem (RCPSP). The multi-mode SoS version exploits the fact that using a state-of-the art solver a small mixed integer linear programming problem (MILP) or a large linear programming problem (LP) can be solved within reasonable time. In order to illustrate the viability of the pre-optimized starting repertoire we present computational results for the hardest and largest MMLIB+ benchmark set developed by Van Peteghem and Vanhoucke [2]. The computational result reveals the fact, that the pre-optimized repertoire drastically increases the efficiency of the problem solvong process.

A. Csébfalvi,

Volume 5, Issue 4 (7-2015)

Abstract

This study has been inspired by the paper "An efficient 3D topology optimization code written in MATLAB” written by Liu and Tovar (2014) demonstrating that SIMP-based three-dimensional (3D) topology optimization of continuum structures can be implemented in 169 lines of MATLAB code. Based on the above paper, we show here that, by simple and easy-to-understand modifications we get a few lines longer code, which is able to solve robust topology optimization problems with uncertain load directions. In the presented worst load direction oriented approach, the varying load directions are handled by quadratic constrains, which describe spherical regions about the nominal loads. The result of the optimization is a robust compliance-minimal volume constrained design, which is invariant to the investigated directional uncertainty. The key element of the robustification is a worstload-direction searching process, which is formulated as a small quadratic programming problem with quadratic constraints. The presented approach is a 3D extension of the robust approach originally developed by Csébfalvi (2014) for 2D continuum structures. In order to demonstrate the viability and efficiency of the extension, we present the model and algorithm with detailed benchmark results for robust topology optimization of 3D continuum structures. It will be demonstrated that the computational cost of the robustification is comparable with its deterministic equivalent because its central element is a standard 3D deterministic multi-load structure optimization problem and the worst-loaddirection searching process is formulated as a significantly smaller quadratically constrained quadratic programming problem, which can be solved efficiently by several different ways.

A. Csébfalvi,

Volume 6, Issue 3 (9-2016)

Abstract

In this paper, a displacement-constrained volume-minimizing topology optimization model is present for two-dimensional continuum problems. The new model is a generalization of the displacement-constrained volume-minimizing model developed by Yi and Sui [1] in which the displacement is constrained in the loading point. In the original model the displacement constraint was formulated as an equality relation, which practically means that the number of “interesting points” may be exactly one. The recent model resolves this weakness replacing the equality constraint with an inequality constraint. From engineering point of view it is a very important result because we can replace the inequality constraint with a set of inequality constraints without any difficulty. The other very important fact, that the modified displacement-oriented model can be extended very easily to handle stress-oriented relations, which will be demonstrated in the forthcoming paper. Naturally, the more general theoretical model needs more sophisticated numerical problem handling method. Therefore, we replaced the original “optimality-criteria-like” solution searching process with a standard nonlinear programming approach which is able to handle linear (nonlinear) objectives with linear (nonlinear) equality (inequality) constrains. The efficiency of the new approach is demonstrated by an example investigated by several authors. The presented example with reproducible numerical results as a benchmark problem may be used for testing the quality of exact and heuristic solution procedures to be developed in the future for displacement-constrained volume-minimization problems.