Prediction Model of Rotor Yarn Quality Based on CNN-LSTM

In the whole textile industry chain, yarn production is one of key links, which has a great impact on quality and clothing products. For long time, been hoping for prediction technology, can accurately predict final indicators according to known conditions such as raw materials processes. CNN-LSTM model deep neural network based assumption that influence processing time series considered. CNN optimizes input eigenvalues through one-dimensional convolution pooling, LSTM matches optimized fiber performance indexes process parameters in sequence excavates their laws, thus realizing goal predicting indexes. The effects index, parameters, kernel pool unit number, layer optimization algorithm accuracy were studied, determined. Experiments data set spinning show mean square error (MSE) strength, Dan Qiang unevenness, evenness total neps lower than linear regression BP network. At same it found greatly influenced by algorithm.