بهینه سازی شبیه سازی چند هدفه برای عملیاتهای ارثمفنج

This paper presents an integrated framework of multi-objective simulation-optimization for optimizing equipment-configurations of earthmoving operations. The earthmoving operations are modeled through simulation and the performances associated with equipment-configurations are evaluated in terms of multiple attribute utility reflecting the preference of decision-makers to multiple criteria. A modified two-stage ranking-selection procedure, a statistical method, is equipped to help compare the alternatives that have random performances and thus reduce unnecessary number of simulation replications. In addition, particle swarm optimization is incorporated to search for the potential equipment-configurations to be examined through simulation, thus speeding up the evaluation process and avoiding exhaustive simulation experiments of all the alternatives. The architecture of the integrated framework is developed. A computational example is provided to justify the proposed methodology. The study will provide an alternative means to help plan earthmoving operations by considering multiple criteria and combining multiple methodologies.

Earthmoving operations are commonly involved in construction engineering and are often performed under conditions that may give rise to uncertainty and randomness. In most cases, earthmoving operations need to be completed within deadline and limited cost. Various methodologies have been proposed to help plan earthmoving operations [1], [2] and [3]. Simulation is one of the methodologies that can be applied to analyze earthmoving operations by modeling uncertainties and randomness. CYCLONE [4] and STRBOSCOPE [5] are the commonly used simulation tools specified for construction. In consideration of the drawback of simulation that requires exhaustive experiments of all the alternative inputs to achieve “What if” or sensitivity analysis, combination of simulation and optimization, namely simulation–optimization, has been proposed to answer not only “What if” but also “How to” questions [6], which is becoming a mainstream in the simulation field [7]. Some efforts have been made with respect to simulation–optimization in construction. Abourizk and Shi [8] and Shi and Abourizk [9] considered constraints on inputs to guide simulation experiments for seeking near-optimal resource quantities. Hegazy and Kassab [10], Cheng and Feng [11], Marzouk and Moselhi [3] and [12] and Cheng et al. [13] proposed simulation–optimization by combining genetic algorithm (GA) with CYCLONE or other simulation techniques. Most of these construction-specified simulation–optimizations focused on single objective optimization rather than multiple criteria. The simulation–optimization of Marzouk and Moselhi [3] and [12] considered multi-objectives for selecting near-optimal fleet configuration for earthmoving operations, but could not select any potential combinations of various types of equipment in a fleet. Moreover, these simulation–optimizations did not address the randomness in the performances produced from simulation, so that the performance differences related to stochastic natures or input variances could not be differentiated when ranking the alternatives. In this study a framework of multi-objective simulation–optimization (MOSO) for optimizing equipment-configurations of earthmoving operations is proposed by integrating an activity object-oriented simulation (AOOS) [14], multiple attribute utility (MAU) theory [15], a statistical approach like the two-stage ranking and selection (R&S) procedure [16] and particle swarm optimization (PSO) algorithm [17] and [37]. The MAU theory is applied to evaluate the performances generated through simulation by considering multiple criteria and the preference of decision-makers. A modified two-stage R&S method will be incorporated to handle randomness in the performances when ranking and selecting best alternative equipment-configurations. Reasonable number of simulation replications for examining each alternative will be determined during two-stage R&S, and thus reducing unnecessary simulation experiments. PSO that is similar to Genetic Algorithm (GA) is used to help search for potential alternatives so as to avoid exhaustive simulation experiments of all available equipment-configurations, enhancing the efficiency of the proposed methodology.

The study proposes an integrated framework for a multi-objective simulation–optimization (MOSO) for determining optimal equipment-configurations of earthmoving operations. The proposed framework combines multiple methodologies such as simulation, multiple attribute utility (MAU), statistics and optimization. An existing simulation platform AOOS is adopted to model an earthmoving operation. The multiple performances such as project duration and total cost can be obtained by conducting simulation experiments and transformed to a total attribute utility that can reflect multiple criteria and the preference of decision-makers. A statistical method, i.e., two-stage ranking and selection (R&S), is incorporated to compare the performances with regard to the differentiation between the random variance and the variance associated with different inputs. Modification on the two-stage R&S is made by considering a subset to include higher potential alternatives that need to be examined with more simulation replications, and thus reducing total number of simulation replications. Particle swarm optimization (PSO) is adopted to help research for potential alternatives and thus avoid exhaustive simulation experiments of all the alternatives as normal simulation does. The architecture of the integrated framework is developed. Transformation between the particle-represented equipment-configurations and the inputs for simulation experiments is built. The experimental analyses have demonstrated the effectiveness of the proposed MOSO in determining optimal equipment-configuration for an earthmoving operation. The efficiency of the proposed EOSO is also justified through analysis of the total number of simulation replications. The study is expected to provide an alternative means to improve planning of earthmoving operations by combining multiple techniques, and thus overcome some drawbacks of the existing methods. Further study will focus on analysis of other parameters for the proposed MOSO and considering optimization of other qualitative variables for earthmoving operations.