The focus on digitalization and how to reap the benefits from it, is increasing in the oil and gas industry. To make this happen subsea by using machine learning and neural networks on mechanical products, we need to understand how we can digitize them.
In this paper we will show how we can generate a true digital twin that could enable the step change in how we monitor and understand the mechanical products placed out of reach for normal preventive maintenance.
We look at the industry’s current information philosophy regarding mechanical products; from idea to development and testing to installation and operation. Where the products are currently designed to end up with a nominal 3D model, test components and the verification that they are fit for purpose is made through a test scope simulating certain scenarios defined by the industry. The products are then installed, and their condition is assumed to be fine until evidence of the opposite is apparent, and the time window to perform mitigative intervention is gone.
Creating true digital twins of mechanical products will require more data. In recent tests we have focused on generating data points to understand the response/behaviour of the products in multiple scenarios and utilized this data to describe the behaviour accurately and numerically; from this we can generate a digital representation of what state the products are in.
The results are based on recent work in developing a digital twin for mechanical products. In this work we have proven through documented testing and quantitative analysis that we can generate a validated numerical model for mechanical products. This will form the basis for understanding the state of the products and predict when intervention/maintenance is imminent for the operator. Using this model and method, condition monitoring of the mechanical products can now be enabled with relative few datapoints extracted in time intervals, and through the use of its numerical representation, establish an actual condition history of the product itself.
Further this model would enable a deeper understanding of the actual operational effects, to which the product is exposed during its lifetime, leading to more precise and cost-efficient industry requirements and system knowledge.