Energy Consumption Prediction in 3D Printing Using Enhanced Hybrid Deep Learning Models

Abstract

The 3D printing method is characterized as an energy-intensive technology with a major negative influence on the ecosystem and sustainability. Since then, both business and academics have turned their attention to the problem of 3D printing energy usage modelling, prediction, and optimization. However, prediction of energy consumption is a common issue in 3D printing. Thus, this research aims to forecast the energy consumption in 3D printing employing a novel Convolutional Neural Network with Long-Short Term Memory (CNN-LSTM). The developed model has five stages: data collection, data augmentation, pre-processing, feature extraction, and prediction. In data collection, Ender-3 pro 3D printer data were collected; whereas, in pre-processing, data normalization was performed by min-max normalization and missing data imputations was carried out using Mean or median. Moreover, feature extraction was done using Principal Component Analysis (PCA) method. Furthermore, in prediction phase, CNN-LSTM is employed for forecasting the energy usage. Moreover, the model’s performance has been evaluated in regards to Mean Square Error (MSE), correlation, Root-MSE (RMSE), Normalized-MSE (NMSE), and Mean-Absolute Error (MAE). Furthermore, comparison has been performed to identify the effectiveness of the presented model over other methods.