Capacity Prediction, Cost Reduction and Efficiency Improvement Optimization for Mold Enterprises Based on LSTM-RF Ensemble Learning

Abstract

To address challenges such as significant capacity fluctuations, low scheduling efficiency, and crude cost control in plastic mold manufacturing facilities, this study proposes an integrated forecasting framework that combines Lonag Short-Term Memory (LSTM) networks with Random Forest (RF) models, integrating greedy scheduling with an economic quantification model to construct a full-chain decision-making system comprising "forecasting–scheduling–evaluation". Based on the multi-dimensional production data of the past 365 days, lagged feature engineering and periodic feature engineering are used to build a dual-model system for rolling forecasts of equipment load rates, energy consumption and work order volumes; the scheduling process is optimised by a priority-weighted rule; and cost reduction benefits are quantified in four dimensions: labour, materials, energy consumption and maintenance. Based on the experiment, the integrated model has achieved a Mean Absolute Error (MAE) of 0.0324 and an R² value of 0.8916 for load rate forecasting; after scheduling optimisation, the idle rate has dropped by 17.82 percentage points, and a total monthly cost saving of about CNY 196,000 has been realised.