The study was carried out to address the difficulty of identifying energy inefficiencies in contexts of industrial production. In pursuit of lower energy usage and improved sustainability, the study of significant machine-level energy data in manufacturing is vital. The problem was found to be the lack of automated techniques for processing, analysing, and deriving meaningful insights from time-stamped energy data. This challenge is already critical in today’s manufacturing industry, as rising energy costs and stringent sustainability targets drive an urgent need for solutions that can turn raw machine logs into actionable decisions.

The science of manufacturing engineering technology has been explored to find a solution. With minimal overhead resources and meetings, the project sought to build a completely automated analytic pipeline capable of converting raw machine data into interpretable metrics, thereby enabling organisations to identify wasteful standby patterns and support energy-optimisation decisions.

To do this, a Python modular data pipeline was built, strategically capable of processing billions of data points each day. The process started with cleaning and formatting unprocessed CSV files, then segmented by machine to create quality reports verifying data integrity. Standardised time intervals and a wide range of attributes were developed from the resampling procedure encompassing time-of-day patterns, status use, and prolonged standby times. Patterns and seasonal elements from the data were obtained by means of time series decomposition. The features were further scaled, and dimensionality dropped.

The results showed that the pipeline efficiently handled large and irregular datasets and that it was possible to spot machines showing unusually high energy consumption during non-productive intervals. Individual deviations were tracked, and machines were assigned behavior-based clusters. At times, over half of a machine's total energy use occurred outside of output, suggesting great room for improvement.

Designed for industrial energy optimization, the pipeline leverages AI and machine learning techniques to run fully autonomously at scale. It converts raw machine logs into interpretable metrics, identifying wasteful standby patterns and informing optimization decisions around scheduling, maintenance and equipment upgrades. Further work will explore forecasting models to predict future inefficiencies and extend the ML framework to real-time energy control.