Hybrid Models for the Design of ML-enabled Machinery for Sustainable Offshore and Subsea Well Service Applications

The traditional well service machinery is manually operated and requires human intervention for optimal performance. The design of this machinery tends to be conservative and leads to bulky and heavy units being deployed offshore. In the recent decade, there has been rapid development and widespread adoption of digital twins and artificial intelligence tools in the product life cycle in many areas. This presents an opportunity for the offshore drilling industry to adopt these new advanced tools to achieve more optimal design and operations such that costs and, in particular, waste can be significantly reduced. It is essential for the oil and gas industry, where waste management is crucial. As the industry is a major water and energy resource consumer, it is subject to ever-increasing stringent environmental standards. O&G companies must rethink their drilling and production activities and reduce their environmental footprint. Therefore, maximum production efficiency and minimal waste generation are top priorities to stay competitive. In this PhD project, hybrid modelling techniques will be developed to support the design of machine learning-enable machinery in the offshore industry. The hybrid model is a model that combines a digital twin with a model based on measurements collected from the field. The digital twin combines models created using CAE, such as the finite element method, computational fluid dynamics and structural reliability models. The measurement-based models consist of measurements collected from the field and fitted using regression techniques. The combination of these two models will allow the combined model to (i) predict events more accurately, as the digital twin is correct with real measurements, and (ii) predict events outside the measured data set that the digital twin offers. This combined model will allow a greater understanding of the product's lifecycles during design, deployment, and maintenance.

Heavy machinery units installed on offshore platforms are supposed to be robust and reliable. Additionally, due to the high mobilisation costs offshore, it is not cost-efficient to perform sufficient services and tests too often. A new and advanced methodology for analysing and converting mechanical tools from 3D design objects to interactable dynamic objects in a resource-efficient way needs to be researched and developed. In this PhD project, a hybrid model consists of the measurements part, and a digital model of the physical machine is proposed. The measurement-based model collects and proceeds data. The digital model is built based on the computational models established using state-of-art methods such as FEA, CFD, MBSE, etc. An example of the way how the hybrid model can be utilised is displayed in the figure below using the DWS WSP 725 CC Well Service Pump. The physical machine (Well Service Pump) that is installed offshore is equipped with different sensors, i.e., motor, gear, pumps, utility etc. The sensors send data with information about the work condition of the machine to the hybrid model. The hybrid model allows one to carry out the tests on the digital model and apply the results to the physical machine. The digital model is called the digital twin because it is an identical virtual copy of the physical. This allows remote testing without going offshore, which saves transport costs. Once the analysis is performed, artificial intelligence using machine learning should be able to differentiate and choose the data that can be applied to the physical unit. A complete system containing these features can be used to research and test new approaches to visualising mechanical simulations.