Accurate and reliable information about the temperature of the synchronous generators excitation winding hot spot is necessary to determine the dynamic limit caused by excitation winding overheating in the PQ diagram....Accurate and reliable information about the temperature of the synchronous generators excitation winding hot spot is necessary to determine the dynamic limit caused by excitation winding overheating in the PQ diagram. For good estimation of a position and the hot spot temperature it is decided to mount 19 temperature probes on one pole of the 6-pole, 400 kVA. 50 llz synchronous generator. Due to a large number of the probes and because the probes should be glued with the metal epoxy it was assumed that mounting of the probes will disrupt the temperature field of the excitation winding. To get the answer to this question the excitation winding resistance was measured betbre and after mounting the probes, in a hot and a cold state. Temperature rise can be estimated if the resistance ratio in the hot and the cold state is known. The paper also addresses the analysis of the measurement accuracy. The result shows that, there is no significant influence on the temperature when mounting the 19 temperature probes which covered 10% of the pole excitation winding surface.展开更多
热误差是影响高精密数控机床加工精度的重要因素。为了提高机床加工精度和性能,减少机床运行中产生的热误差,文章提出一种基于热图像的灰狼优化算法(grey wolf optimization algorithm,GWOA)和双向长短期记忆神经网络(bidirectional lon...热误差是影响高精密数控机床加工精度的重要因素。为了提高机床加工精度和性能,减少机床运行中产生的热误差,文章提出一种基于热图像的灰狼优化算法(grey wolf optimization algorithm,GWOA)和双向长短期记忆神经网络(bidirectional long short-term memory,BiLSTM)混合的热误差预测模型。首先,采用热成像仪获取机床主轴区域的温度场信息;其次,利用DBSCAN聚类(density-based spatial clustering of applications with noise)算法和相关系数法筛选出温度敏感点;然后,通过模拟灰狼群体捕食行为,在参数空间中进行搜索以找到BiLSTM所需的最优参数;最后,使用获得的机床温度敏感点和热位移数据进行热误差预测,并在试验机床上进行验证。实验结果表明,使用GWOA优化BiLSTM神经网络的预测模型相比BiLSTM神经网络预测模型的均方根误差(root mean square error,RMSE)和平均绝对误差(mean absolute error,MAE)分别减小了约0.5180、0.3823μm,决定系数R^(2)提升了0.0578。与BiLSTM神经网络模型相比,利用GWOA优化后的模型具有更加优良的预测性能。展开更多
文摘Accurate and reliable information about the temperature of the synchronous generators excitation winding hot spot is necessary to determine the dynamic limit caused by excitation winding overheating in the PQ diagram. For good estimation of a position and the hot spot temperature it is decided to mount 19 temperature probes on one pole of the 6-pole, 400 kVA. 50 llz synchronous generator. Due to a large number of the probes and because the probes should be glued with the metal epoxy it was assumed that mounting of the probes will disrupt the temperature field of the excitation winding. To get the answer to this question the excitation winding resistance was measured betbre and after mounting the probes, in a hot and a cold state. Temperature rise can be estimated if the resistance ratio in the hot and the cold state is known. The paper also addresses the analysis of the measurement accuracy. The result shows that, there is no significant influence on the temperature when mounting the 19 temperature probes which covered 10% of the pole excitation winding surface.
文摘热误差是影响高精密数控机床加工精度的重要因素。为了提高机床加工精度和性能,减少机床运行中产生的热误差,文章提出一种基于热图像的灰狼优化算法(grey wolf optimization algorithm,GWOA)和双向长短期记忆神经网络(bidirectional long short-term memory,BiLSTM)混合的热误差预测模型。首先,采用热成像仪获取机床主轴区域的温度场信息;其次,利用DBSCAN聚类(density-based spatial clustering of applications with noise)算法和相关系数法筛选出温度敏感点;然后,通过模拟灰狼群体捕食行为,在参数空间中进行搜索以找到BiLSTM所需的最优参数;最后,使用获得的机床温度敏感点和热位移数据进行热误差预测,并在试验机床上进行验证。实验结果表明,使用GWOA优化BiLSTM神经网络的预测模型相比BiLSTM神经网络预测模型的均方根误差(root mean square error,RMSE)和平均绝对误差(mean absolute error,MAE)分别减小了约0.5180、0.3823μm,决定系数R^(2)提升了0.0578。与BiLSTM神经网络模型相比,利用GWOA优化后的模型具有更加优良的预测性能。
