Application of a Data-Driven Dynamic Process Modelling Framework to Improve the Condition Monitoring of a Liquefied Petroleum Gas Refrigeration Unit

With the fast-growing digitalisation of maintenance procedures in gas processing units, there has been a recent demand for data-driven condition monitoring (CM) tools that could competently diagnose the unit’s defective behaviour at off-design conditions and subsequently execute necessary adjusting measures. This demand mainly stems from the incapability of traditional process simulator packages in predicting the actual dynamic performance of such units, which not only leads to corresponding CM strategies becoming inaccurate in response, but, on a larger scale, makes the unit and its dependent downstream processes operate inefficiently at high maintenance costs. Towards addressing the noted matter for a practical case, this investigation proposes a data-driven CM framework for a malfunctioning refrigeration unit in a reference liquefied petroleum gas (LPG) plant, aiming to cost-effectively predict and resolve the unit’s faulty behaviour in a real-time sense. The framework is developed by customising the unit dynamic process simulators – through the actual history of unit operation attained from built-in measurement tools – and then coupling the simulator’s output to the plant control system for predetermined optimisation tasks. Considering the high adaptivity level of the proposed framework, it can be conveniently used as a supplemental tool for enhancing CM procedures in relevant processing plants.