Neural Network-Assisted Forecasting and Scheduling for Energy- Aware Manufacturing

Abstract

The focus of this research is on the energy consumption of manufacturing processes. In response to this concern,
industries have been urged to adopt advanced strategies for forecasting and scheduling production to optimize the
consumption of resources. This work includes the design of a system that implements real time forecasting and
scheduling within the context of energy aware manufacturing using neural network techniques. Based on available
historical data, a Multi-Layer Perceptron (MLP) model is developed that accurately predicts the energy requirement
for given tasks, feeding those predictions to the scheduling module which dynamically reconfigures job order to lessen
energy expenditure while adhering to completion deadlines. Comprehensive simulations are performed to ascertain
the efficacy of the hybrid model in diverse manufacturing environments. Findings highlight that, in comparison to
traditional models, there was a marked improvement in energy efficiency (up to 18% reduction) while maintaining
the same levels of throughput. The benchmarks for the proposed framework included rule-based and heuristic
schedulers and various power function precision formulas, all of which, along with statistical estimations confirmed
the postulated claims of the research regarding efficiency and accuracy. This research significantly contributes to the
body of knowledge focusing on the implications of artificial intelligence in manufacturing systems, laying a solid
groundwork for research directed toward planning with energy considerations in manufacturing systems. Primary
attention in future works will shift toward practical implementation of the intelligent control system along with
renewable energy forecast integration and enhancing adaptability to varying production schedules.