A Benchmarking-Based Framework for Energy Performance Variation Screening in Manufacturing Systems

Abstract

Energy data are increasingly available in manufacturing systems, but valid interpretation depends on production context that is often incomplete. This thesis develops a benchmarking-based framework for screening and classifying within-resource energy-performance variation under limited data conditions. The framework is applied to an industrial case using three datasets provided by Volvo Trucks: 30-minute device-level energy records, stop logs, and stop-detail records. The method integrates the datasets at a common resource–time-window level, constructs comparable active-like windows, establishes a local benchmark from historically lower-energy observations, and screens high benchmark-gap windows using the available stop information. Energy use per 30-minute window is treated as a screening variable rather than an output-normalised efficiency indicator. The application identifies distinct variation profiles across resources: some resources show larger relative gaps, whereas others show greater accumulated recorded energy above the local benchmark. Recorded stop events provide candidate explanations for only part of the flagged windows, demonstrating that richer data on output, product type, cycle time, shift, and machine state are required for causal diagnosis. The framework therefore supports transparent screening and prioritisation, but it does not by itself establish energy inefficiency, quantify savings potential, or confirm root causes.