基于注塑机螺杆位置与压力曲线的注射成型过程监测方法

提出基于注塑机螺杆位置与压力信号,采用非线性数据降维方法提取成型特征,建立原材料、模具温度与制品成型质量之间的神经网络模型,实现注射成型过程的故障监测和质量预测。试验结果表明：与未降维的数据和主成分分析等线性降维方法相比,拉普拉斯映射法和扩散系数图法相结合的非线性降维方法可从螺杆位置与压力信号中判断制品原材料、模具温度的异常及预测制品成型质量,在验证集和测试集均获得了高于0.92的回归系数,所提出的方法提取了原始曲线数据内在的高维度、非线性、强耦合的数据关系,有效进行了数据降维,提高了故障监测与质量预测模型的精度与效率。

Monitoring method of injection moulding process based on screw position and pressure curve of injection molding machine

Based on the screw position and pressure signals of injection molding machine,the moulding characteristics were extracted by using the nonlinear data dimensionality reduction method,and the neural network model among raw material,mould temperature,and moulding quality was established,and a method of fault monitoring and quality prediction in injection moulding process was proposed.The test results showed that the nonlinear dimensionality reduction method combined with Laplace eigenmaps and diffusion coefficient graph could judge the abnormal material and mould temperature and predict product quality from screw position and pressure signals,compared with linear dimensionality reduction methods such as non-dimensionality reduction data and principal component analysis.The regression coefficients were higher than 0.92 in both validation and test sets,and the proposed method extracted the high-dimensional,nonlinear,and strongly coupling data relations inherent in the original curve data,effectively reduced the data dimension,and improved the accuracy and efficiency of the fault monitoring and quality prediction model.