ABSTRACT

A hybrid approach through combining genetic algorithm (GA) and adaptive neuro-fuzzy inference system (ANFIS) for modeling the correlation of laser beam cutting (LBC) parameters and enhancing the quality performance characteristics of machined Hastelloy C276 is emphasized. The LBC experiments are performed by considering gas pressure (GP), cutting speed (CS), pulse energy (PE) and stand-off distance (SOD) as input parameters. The output responses are material removal rate (MRR), kerf taper (KT) and surface roughness (R_a) for the present investigation. The optimal ANFIS training variables are obtained through GA. The training, testing errors, and statistical validation parameter results exposed that the ANFIS learned by GA is outperformed in forecasting LBC responses. In addition, to obtain the optimal combinations of LBC parameters, the multi-response optimization based on maximizing MRR and minimizing KT and R_a was performed using a trained ANFIS network coupled with a whale optimization algorithm (WOA). The responses such as MRR of 236.98 mg/min, KT of 1.135° and R_a of 1.109 µm are forecasted for the optimum cutting conditions: GP of 3 bar, CS of 319.8 mm/min, PE of 5.93 J and SOD of 2.97 mm, respectively. Furthermore, the WOA predicted results are validated by conducting confirmatory experiments.

摘要

强调了一种通过结合遗传算法 (GA) 和自适应神经模糊推理系统 (ANFIS) 的混合方法，对激光束切割 (LBC) 参数的相关性进行建模，并提高加工哈氏合金 C276 的质量性能特征。LBC 实验是通过考虑气体压力 ( GP )、切割速度 ( CS )、脉冲能量 ( PE ) 和间隔距离 ( SOD ) 作为输入参数来进行的。输出响应是材料去除率 ( MRR )、切口锥度 ( KT ) 和表面粗糙度 ( R _a) 用于目前的调查。最优的 ANFIS 训练变量是通过 GA 获得的。训练、测试错误和统计验证参数结果表明 GA 学习的 ANFIS 在预测 LBC 响应方面表现优于。此外，为了获得 LBC 参数的最佳组合，使用训练有素的 ANFIS 网络结合鲸鱼优化算法 (WOA) 执行基于最大化MRR和最小化KT和R _a的多响应优化。对于最佳切削条件，预测了诸如MRR为 236.98 mg/min、KT为 1.135° 和R _a为 1.109 µm 等响应：GP为 3 bar，CS为 319.8 mm/min，PE为 5.93 J，SOD为 2.97 mm。此外，通过进行验证性实验验证了 WOA 预测结果。