摘要
针对铣削加工中刀具磨损预测精度不足的问题,为提高预测准确率并实现有效的状态监测,提出一种基于极光优化算法(PLO)优化的Transformer与长短时记忆神经网络(LSTM)融合预测方法。对振动、切削力等多传感器信号进行时域、频域及时频域的多域特征提取,构建与磨损状态相关的特征集。利用PLO算法对Transformer模型的关键超参数进行自动优化,以克服手动调参效率低的问题。将优化后的Transformer与LSTM网络相结合,利用Transformer的全局注意力机制捕捉长程依赖关系,并结合LSTM处理序列数据的优势,构建PLO-Transformer-LSTM预测模型。最后,基于PHM2010公开数据集的铣刀全寿命周期振动数据开展实验验证。结果表明:该模型预测铣刀磨损量的平均绝对百分比误差(MAPE)为1.2533%,均方根误差(RMSE)为2.4754,与GA-LSTM、BWO-LSTM-AdaBoost等模型相比,所提方法在MAPE和RMSE指标上均表现更优,显著提升了铣刀磨损量的预测精度。
To address the issue of insufficient prediction accuracy for tool wear in milling and to improve prediction accuracy for effective condition monitoring,a hybrid prediction method based on the fusion of Transformer and long short-term memory(LSTM)networks optimized by the polar lights optimization(PLO)algorithm was proposed.Multi-domain features from time,frequency and time-frequency domains were extracted from multi-sensor signals such as vibration and cutting force,to construct a feature set related to wear state.The PLO algorithm was used to automatically optimize the key hyperparameters of the Transformer model,overcoming the inefficiency of manual parameter tuning.The optimized Transformer was combined with an LSTM network,leveraging the global attention mechanism of Transformer to capture long-range dependencies and the strength of LSTM in processing sequential data,and the PLO-Transformer-LSTM prediction model was constructed.Finally,experiments were conducted using the full-life-cycle vibration data of milling tools from the PHM2010 public dataset.Results show that a mean absolute percentage error(MAPE)of 1.2533%and a root mean square error(RMSE)of 2.4754 are achieved by the proposed model for milling tool wear prediction.Compared with models such as GA-LSTM and BWO-LSTM-AdaBoost,the proposed method performs better in both MAPE and RMSE metrics,significantly improving the prediction accuracy of milling tool wear.
作者
龚琛璞
李波
付文杰
GONG Chenpu;LI Bo;FU Wenjie(Institute of Mechanical Engineering,Hubei University of Arts and Science,Xiangyang Hubei 441053,China;XY-HUST Advanced Manufacturing Engineering Research Institute,Xiangyang Hubei 441100,China)
出处
《机床与液压》
北大核心
2026年第5期144-149,共6页
Machine Tool & Hydraulics
基金
湖北省重大科技攻关项目(2023BAA010)。
