To overcome the disadvantages of premature convergence and easy trapping into local optimum solutions, this paper proposes an improved particle swarm optimization algorithm (named NDWPSO algorithm) based on multiple hybrid strategies. Firstly, the elite opposition-based learning method is utilized to initialize the particle position matrix. Secondly, the dynamic inertial weight parameters are given to improve the global search speed in the early iterative phase. Thirdly, a new local optimal jump-out strategy is proposed to overcome the "premature" problem. Finally, the algorithm applies the spiral shrinkage search strategy from the whale optimization algorithm (WOA) and the Differential Evolution (DE) mutation strategy in the later iteration to accelerate the convergence speed. The NDWPSO is further compared with other 8 well-known nature-inspired algorithms (3 PSO variants and 5 other intelligent algorithms) on 23 benchmark test functions and three practical engineering problems. Simulation results prove that the NDWPSO algorithm obtains better results for all 49 sets of data than the other 3 PSO variants. Compared with 5 other intelligent algorithms, the NDWPSO obtains 69.2%, 84.6%, and 84.6% of the best results for the benchmark function ( $${f}_{1}-{f}_{13}$$ ) with 3 kinds of dimensional spaces (Dim = 30,50,100) and 80% of the best optimal solutions for 10 fixed-multimodal benchmark functions. Also, the best design solutions are obtained by NDWPSO for all 3 classical practical engineering problems.
This article constructed a vehicle scheduling problem (VSP) with soft time windows for a certain ore company. VSP is a typical NP-hard problem whose optimal solution can not be obtained in polynomial time, and the basic particle swarm optimization(PSO) algorithm has the obvious shortcoming of premature convergence and stagnation by falling into local optima. Thus, a modified particle swarm optimization (MPSO) was proposed in this paper for the numerical calculation to overcome the characteristics of the optimization problem such as: multiple constraints and NP-hard. The algorithm introduced the “elite reverse” strategy into population initialization, proposed an improved adaptive strategy by combining the subtraction function and “ladder strategy” to adjust inertia weight, and added a “jump out” mechanism to escape local optimal. Thus, the proposed algorithm can realize an accurate and rapid solution of the algorithm’s global optimization. Finally, this article made typical benchmark functions experiment and vehicle scheduling simulation to verify the algorithm performance. The experimental results of typical benchmark functions proved that the search accuracy and performance of the MPSO algorithm are superior to other algorithms: the basic PSO, the improved particle swarm optimization (IPSO), and the chaotic PSO (CPSO). Besides, the MPSO algorithm can improve an ore company’s profit by 48.5–71.8% compared with the basic PSO in the vehicle scheduling simulation.
In this work, the parallel numerical simulations of 3-dimensional (3D) mathematical model for the walking-beam type reheating furnace have been developed and implemented on the graphics processing unit (GPU) architecture. Firstly, the detailed heat transfer processes in the furnace are described and categorized when building the 3D mathematical model. They consist of the radiative heat exchange into the slab, the heat conduction between the stationary beams and the slabs, the heat convection between the gas flow and slab surfaces and the heat conduction inside the slabs. And the proposed 3D mathematical model also accounts for the temperature-dependent material parameters, which is ignored by most published mathematical models. Secondly, the explicit finite difference method is used to discretize the proposed model to a straightforward parallel computation problem. A detailed analysis of the 3D boundary conditions for the proposed model is introduced and presented. The parallel computing problem is realised by programming on GPU via the platform CUDA in Tesla P100. Finally, the proposed model is verified with industry measurements, and the comparison between the GPU-implementation model and the CPU-implementation model is also given to validate the great acceleration. Experimental results prove that the proposed GPU-implementation model declines the computation time from hours to seconds. It is not only orders of magnitude faster but also highly accurate.
Each slab entering to the reheating furnace has an optimal and unique reheating curve. The process of obtaining the optimal reheating curve is to solve the typical Partial differential equations (PDE) constrained optimization problem. Obviously, the solution of optimization problem is determined by both the precision of the mathematical PDE model and the numerical method. Firstly, the more accurate mathematical PDE model, in which some key parameters are reconsidered as temperature-dependent, is built for the reheating furnace. Secondly, the first-optimize-then-discretize approach is introduced to solve this PDE-constrained optimization problem. The analysis of the Fr é chet gradient of the cost functional is given and we can prove the gradient is Lipschitz continuous. Then, an improved conjugate gradient method is proposed to solve this problem. Finally, numerical simulations and experiment examples are given and analyzed. The results can prove the effectiveness of the proposed strategy.
Aiming at the problem of the low accuracy of temperature prediction, a mathematical model for predicting the temperature of a steel billet is developed. For the process of temperature prediction, an improved particle-swarm-optimization algorithm (called XPSO) is developed. XPSO was designed based on a multiple swarm scheme to improve the global search capability and robustness; thus, it can improve the low accuracy of prediction and overcome the problem of easy entrapment into local optima. In the XPSO, the multiple swarm scheme comprises four modified components: (1) the strategy of improving the positional initialization; (2) the mutation strategy for particle swarms; (3) the adjustment strategy of inertia weights; (4) the strategy of jumping out local optima. Based on widely used unimodal, multimodal and composite benchmark functions, the effectiveness of the XPSO algorithm was verified by comparing it with some popular variant PSO algorithms (PSO, IPSO, IPSO2, HPSO, CPSO). Then, the XPSO was applied to predict the temperatures of steel billets based on simulation data sets and measured data sets. Finally, the obtained results show that the XPSO is more accurate than other PSO algorithms and other optimization approaches (WOA, IA, GWO, DE, ABC) for temperature prediction of steel billets.
6 January 2025 in International Journal of Thermal Sciences