Lite Optimisation and Recommender Systems for Maintenance Inventory Management

Abstract

Abstract Objectives/Scope This work aims to develop a systematic methodology for the optimisation of maintenance inventory management, independent of industry or project lifecycle stages (greenfield and brownfield). The objective is to enable operators to integrate the use of ‘lite’ dynamic optimisation and recommender systems into existing or proposed business processes, systems, and ways of working. Additionally, this work aims to explore methodologies for enabling the sustainment and ongoing improvement of such systems within a complex asset management context. Methods, Procedures, Process Leveraging Natural Language Processing (NLP) algorithms to improve the speed and accuracy of identifying and cataloguing interchangeable materials and classifying failure modes from operational maintenance data. These algorithms enable comparative analytics across and within facilities, including reliability modelling, material demand modelling, and production vs maintenance optimisation models. This approach leverages foundational maintenance build analyses outputs such as FMECA, RCM, and cataloguing. Results, Observations, Conclusions Results highlight the ability of these ‘lite’ dynamic optimisation and recommender models to be deployed at speed and to be sustained, whilst enabling significant value gains, across production uptime, maintenance cost, and safety, at low cost (particularly when compared to large software systems). Such systems are also designed to be user and use case centric, enabling them to be dynamic and easy to configure. A recent implementation of our data-driven approach has led to 40% reduction in inventory value while increasing the critical spare materials to enable high asset availability. Novel/Additive Information Advanced text processing algorithms used in conjunction with advanced maintenance and reliability modelling techniques, packaged within ‘lite’ systems. These systems are also live and provide recommendations at a frequency suitable to the operator's business process requirements.