Prediction Model of Rotor Yarn Quality Based on CNN-LSTM

Abstract

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