基于神经网络的液固挤压工艺参数灵敏性分析

将人工神经网络引入液固挤压工艺参数的灵敏性分析中，对难以建立精确数学模型的液固挤压工艺进行建模，通过非线性网络泛化映射，求解输出变量对输入变量的偏导数，得到了工艺参数在每个样本点处的灵敏度值，从而定量地确定了多个非确定性参数共同作用下的灵敏度指标。结果表明，影响液固挤压工艺的参数中，作用最大的为浸渗时间，其次为浇注温度与模具温度，最小为浸渗压力，这与实际情况相符。

Parameter Sensitivity Analysis for Liquid-solid Extruding Process Based on the Neural Network

A model for liquid-solid extruding process is established by introducing an artificial neural network into parameter sensitivity analysis during the liquid-solid extruding process. Sensitivity value at per sample point of the processing parameters can be obtained to quantitatively determine the sensitivity index under co-action of several uncertainty parameter conditions by mapping nonlinear network to resolve partial differential coefficient of output via input based on the model. The results show that the influencing orders on liquid-solid extruding process is infiltrating time, and then pouring temperature and mould temperature, and then infiltrating pressure, which are accordant with experimental ones.