Particle swarm optimisation (PSO)-based tool path planning for 5-axis flank milling accelerated by graphics processing unit (GPU)

Abstract

Multi-axis machining offers higher machining efficiency and superior shaping capability compared to 3-axis machining. Machining error control is a critical issue in 5-axis flank milling of complex geometries and there is still a lack of solutions. Previous studies have shown that optimisation-based tool path planning is a feasible approach to reduction of machining error. However, the error estimation is time-consuming in the optimisation process, thus limiting the practicality of this approach. In this work, we apply graphics processing unit (GPU) computing technology to solve this problem. A particle swarm optimisation (PSO)-based optimisation scheme is developed to generate a series of cutter locations (CLs) that produce a minimised error on the machined surface. The error amount induced by each CL is simultaneously calculated by the parallel processing units of GPU. The PSO search process driven by the aggregated result is effectively accelerated. Test results show that our approach outperforms previous optimisation methods in both solution quality and computation efficiency. This work demonstrates a novel application of GPU on Computer Aided Manufacturing/Computer Numerical Control.

Optimized tool path planning in 5-axis flank machining using electromagnetism-like algorithms

“…The number of interpolation between consecutive cutter locations also relates to the precision. The values of those parameters were chosen as recommended by the previous studies [3][4][5].…”

“…A tool path consists of N cutter locations. A cutter location can be adjusted in 3D space by varying a set of 8 parameter values [3]. To obtain an optimal tool path yielding minimized geometric deviations requires adjustment of all cutter locations simultaneously.…”

“…Previous studies [2][3][4] have shown that meta-heuristic algorithms provide a systematic approach to controlling and reducing the geometrical deviations in 5-axis flank finishing cut of ruled surfaces through optimization of tool path planning. However, the optimization schemes based on ACS and various PSO algorithms failed to produce good search results due to high dimensionality of the solution space.…”

Computer-based management of interactive data transformation systems using Taguchi’s robust parameter design

“…Metaheuristic methods or natural-intelligence-inspired approximation optimisation techniques are then introduced to solve such problems, which can avoid becoming trapped in local optima with poor values. The process repeats until the flock reaches the best destination (Hsieh and Chu 2011). The particle swarm optimisation method (PSO; Kennedy and Eberhart 1995) is an optimisation approach based on stochastic population.…”

Parallelization of the Honeybee Search Algorithm for Object Tracking

“…The research of Tan and Ding [19] proposes that when an SI algorithm is parallelized through the GPU, the implementation must fall in one of the following four categories: (1) naïve parallel model, (2) multiphase parallel model, (3) all-GPU parallel model, or (4) multiswarm parallel model. They identified those models by observing several parallel implementations of SI algorithms [20][21][22][23][24], most of which are based on the Particle Swarm Optimization heuristic. We selected HSA for this study as it differs from other SI algorithms on the following basis: (1) it can be used for real-valued optimization problems, while other heuristics are limited to discrete problems, (2) it is not explicitly designed for route searching problems, and (3) according to our knowledge, the parallelization of HSA has not been proposed or developed for object tracking.…”