Process production process quality prediction model based on LSTM optimized by SSA

Abstract

Abstract Process production in manufacturing industry has the characteristics of strong continuity and complex timing coupling. To solve the problem of gradient explosion or disappearance when using traditional neural network for multi-step prediction, a multi-step time series prediction model based on sparrow search algorithm and long short-term memory network is constructed. The constructed model uses the sparrow search algorithm to optimize the learning rate, the number of nodes in two hidden layers and the number of iterations of the LSTM model to obtain the optimal network. The process index data of a domestic manufacturing enterprise were selected to achieve multi-step prediction, and five indexes were evaluated: mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE) and R-squared coefficient. The result shows that the constructed SSA-LSTM model has the lowest prediction error, the largest R-squared coefficient and more accurate prediction value, which can provide ideas and ways for enterprises to adjust production plans in advance.