An LJDRNN-based efficient energy intensity prediction in carbon fiber composite material manufacturing process

Abstract Carbon Fibers (CFs) are usually derived from a polyacrylonitrile precursor during composite fiber production. Due to high-temperature needs, the CF composite manufacturing process is considerably energy-consuming and costly. Hence, energy intensity prediction is required. Energy intensity is predicted by the prevailing research techniques; however, it is not predicted accurately. Thus, a Limit Jordan Deep Recurrent Neural Network (LJDRNN)-centric energy intensity prediction was proposed for accurately predicting the energy intensity. Primarily, the material values and processes are detected. Next, for the determined values, pre-processing is done; in addition, the components are extracted from the input values. Then, by employing the Linear Interpolated Honey Badger Optimization (LIHBO), the optimal components are selected. Next, LJDRNN predicts the energy intensity by deploying the optimal components. The proposed LJDRNN achieved an accuracy of 98.32%, outperforming the JRNN (92.10%), RNN (87%), ANN (78%), and CNN (86%), thus demonstrating its superiority in energy intensity prediction. Grounded on the performance measures, the proposed technique’s performance is weighed against the prevailing research techniques where the proposed system attained enhanced performance when analogized to the prevailing methodologies. This study is of high significance to CF producers, offering a robust tool to predict and manage energy consumption effectively. By enabling more precise energy intensity forecasting, the proposed method supports producers in optimizing their manufacturing processes, reducing energy costs, and aligning with sustainable production goals, ultimately driving greater operational efficiency and competitiveness in the CF industry.