部分稳定氧化锆(Partially stabilized zirconia,PSZ)陶瓷因其优越的性能在航空航天工业等领域有广泛的应用。表面粗糙度是评价PSZ陶瓷磨削加工水平的关键指标,为了降低磨削表面粗糙度的预测误差,提出了一种基于相关性分析与卷积-双向...部分稳定氧化锆(Partially stabilized zirconia,PSZ)陶瓷因其优越的性能在航空航天工业等领域有广泛的应用。表面粗糙度是评价PSZ陶瓷磨削加工水平的关键指标,为了降低磨削表面粗糙度的预测误差,提出了一种基于相关性分析与卷积-双向长短期记忆神经网络(Convolution-bidirectional long short term memory neural network,CNN-BiLSTM)的PSZ陶瓷磨削表面粗糙度声发射预测模型。通过分析磨削声发射信号特征值与磨削表面粗糙度值之间相关性,筛选出磨削声发射信号与磨削表面粗糙度之间的最相关频段和特征矩阵,作为CNN-BiLSTM神经网络的输入参数以降低磨削表面粗糙度声发射预测的误差。研究结果表明,基于相关性分析与CNN-BiLSTM神经网络的PSZ陶瓷磨削表面粗糙度的平均预测误差低于3.92%。展开更多
The theoretical model of axial ultrasonic vibration grinding force is built on the basis of a mathematical model of cutting deforming force deduced from the assumptions of thickness of the undeformed debris under Rayl...The theoretical model of axial ultrasonic vibration grinding force is built on the basis of a mathematical model of cutting deforming force deduced from the assumptions of thickness of the undeformed debris under Rayleigh distribution and a mathematical model of friction based on the theoretical analysis of relative sliding velocity of abrasive and workpiece. Then, the coefficients of the ultrasonic vibration grinding force model are calculated through analysis of nonlinear regression of the theoretical model by using MATLAB, and the law of influence of grinding depth, workpiece speed, frequency and amplitude of the mill on the grinding force is summarized after applying the model to analyze the ultrasonic grinding force. The result of the above-mentioned law shows that the grinding force decreases as frequency and amplitude increase, while increases as grinding depth and workpiece speed increase; the maximum relative error of prediction and experimental values of the normal grinding force is 11.47% and its average relative error is 5.41%; the maximum relative error of the tangential grinding force is 10.14% and its average relative error is 4.29%. The result of employing regression equation to predict ultrasonic grinding force approximates to the experimental data, therefore the accuracy and reliability of the model is verified.展开更多
文摘部分稳定氧化锆(Partially stabilized zirconia,PSZ)陶瓷因其优越的性能在航空航天工业等领域有广泛的应用。表面粗糙度是评价PSZ陶瓷磨削加工水平的关键指标,为了降低磨削表面粗糙度的预测误差,提出了一种基于相关性分析与卷积-双向长短期记忆神经网络(Convolution-bidirectional long short term memory neural network,CNN-BiLSTM)的PSZ陶瓷磨削表面粗糙度声发射预测模型。通过分析磨削声发射信号特征值与磨削表面粗糙度值之间相关性,筛选出磨削声发射信号与磨削表面粗糙度之间的最相关频段和特征矩阵,作为CNN-BiLSTM神经网络的输入参数以降低磨削表面粗糙度声发射预测的误差。研究结果表明,基于相关性分析与CNN-BiLSTM神经网络的PSZ陶瓷磨削表面粗糙度的平均预测误差低于3.92%。
基金Project(51275530)supported by the National Natural Science Foundation of China
文摘The theoretical model of axial ultrasonic vibration grinding force is built on the basis of a mathematical model of cutting deforming force deduced from the assumptions of thickness of the undeformed debris under Rayleigh distribution and a mathematical model of friction based on the theoretical analysis of relative sliding velocity of abrasive and workpiece. Then, the coefficients of the ultrasonic vibration grinding force model are calculated through analysis of nonlinear regression of the theoretical model by using MATLAB, and the law of influence of grinding depth, workpiece speed, frequency and amplitude of the mill on the grinding force is summarized after applying the model to analyze the ultrasonic grinding force. The result of the above-mentioned law shows that the grinding force decreases as frequency and amplitude increase, while increases as grinding depth and workpiece speed increase; the maximum relative error of prediction and experimental values of the normal grinding force is 11.47% and its average relative error is 5.41%; the maximum relative error of the tangential grinding force is 10.14% and its average relative error is 4.29%. The result of employing regression equation to predict ultrasonic grinding force approximates to the experimental data, therefore the accuracy and reliability of the model is verified.
