Hyperdimensional Computing for Sustainable Manufacturing: An Initial Assessment

TL;DR

AI-driven sensing in smart manufacturing faces energy challenges that can erode sustainability gains. This paper benchmarks Hyperdimensional Computing (HDC) against conventional AI models on an in-situ CNC machining testbed to predict geometric deviation, focusing on accuracy, speed, and energy. Results show HDC matches top-model accuracy while dramatically reducing training and inference time and energy by up to orders of magnitude. The work highlights substantial potential energy and CO2 savings in real-time manufacturing and points to future extensions across processes and closed-loop control.

Abstract

Smart manufacturing can significantly improve efficiency and reduce energy consumption, yet the energy demands of AI models may offset these gains. This study utilizes in-situ sensing-based prediction of geometric quality in smart machining to compare the energy consumption, accuracy, and speed of common AI models. HyperDimensional Computing (HDC) is introduced as an alternative, achieving accuracy comparable to conventional models while drastically reducing energy consumption, 200$\times$ for training and 175 to 1000$\times$ for inference. Furthermore, HDC reduces training times by 200$\times$ and inference times by 300 to 600$\times$, showcasing its potential for energy-efficient smart manufacturing.

Figures (3)

• Figure 1: Flowchart of the overall approach adopted in this paper.
• Figure 2: Comparison of HDC to other deep learning models in terms of accuracy, precision, recall, and F1-Score for a) 25.4 mm Diameter feature and b) 2.54 mm radius feature.
• Figure 3: Comparison of HDC to other models. Increase in time and energy efficiency relative to HDC is shown next to each bar. (a) Average total training time and speedup (b) Average total training energy usage (c) Average total inference time and (d) Average total inference energy usage.