(Not Quite) Unique Complexity of Piping Simulations: Neural Network Insights in Flow Assurance

Evgenii Lykov

Expert Analysis by Evgenii Lykov, Software Engineer

Although the intricacies of pipe system simulations might seem peculiar to this domain at first glance, they reveal the rich field of similarities and relationships with other scientific and engineering fields under scrutiny. This convergence does not just represent an academic curiosity but instead acts as the core foundation for how these systems are understood, engineered, and optimised.

Evgenii Lykov, a well-known software engineer in this field, investigates the intersections of neural networks and flow assurance in pipe systems. His analysis offers profound insights into the potential applications, challenges, and transformative possibilities that arise from this interdisciplinary approach.

The correspondence between electromagnetism and hydrodynamics was first raised by Lames Clerk Maxwell in his pioneering investigations. Later, the analogy was established between the flow of an electrical current and that of a liquid, which builds a conceptual link between hydraulic engineering and electronic theory on one hand and fluid dynamics in piping systems on the other hand. As a result, engineers who use this approach can devise innovative solutions for difficult problems.

Exploring the detailed landscape of piping simulations reveals their impressive, but not-quite-unique complexity, and also their wide range of applications. In this article, we will examine some of them: the similarities of piping systems to neural networks, which suggests that the adoption of interdisciplinary methods in engineering could lead to groundbreaking developments; furthermore, we will dive into the analogy between joints and neurons or pipes with connections, which implies that neural network approaches might play a role in better pipe system simulations or optimisations; finally, we will try to unveil different applications of piping systems, advocating for AI integration into typical engineering challenges like developing novel simulators for pipes.

Non-Obvious Similarity: Neural Networks in Flow Assurance 

Neural network methods offer an interesting opportunity for managing flow assurance in piping systems that is unique among existing techniques. The interdisciplinary union of these two scientific areas allows for combining their fundamental principles, and due to this fact, many similar aspects are able to be useful in the development of new technologies for the optimisation of flow management processes appear. Having noted the connection between the two fields, now let’s move on to analyse this convergence from various perspectives.

Understanding the Fundamentals

Flow assurance within the piping systems means that the passage of the fluids is uninterrupted, reliable, and safe so that none of them is compromised. Four significant parameters impacting this process comprise pressure, flow rate, velocity, and resistance in the pipeline. The usual way of managing these parameters is to divide the pipes into segments and then perform downstream calculations to forecast how fluids will behave under different conditions.

However, backward calculations—attempting to deduce system characteristics from desired outcomes—are notably less effective and often impractical because of the complexity and non-linearity of the system. Precise tuning of parameters is critical and requires a deep understanding of the system’s behaviour and an ability to predict how changes in one part of the system will affect the whole.

Drawing Parallels

Layering up and using weight connections, artificial intelligence can effectively implement a neural network in which the basic unit is a neuron; neurons with inputs, weights, and outputs are linked together to form layers. This structure (and method )bear a non-obvious yet profound similarity to the challenges faced in flow assurance.

In neural network training, forward calculation processes input data through various layers of the network to generate predictions, leveraging the architecture of the network to model complex relationships. Similarly, back-allocation in flow assurance involves working with known output data, like fluid delivery at the end of a pipeline, and inferring backwards to understand how inputs must have behaved throughout the system to achieve those outputs.

Neural Network-Inspired Piping Networks

Exploring further, piping networks that function and operate according to the principles of neural networks might be conceived as follows:

  • Joints as neurons: All junctions of a hydraulic network can be treated as neurons. Like neurons that transfer signals from one neuron to another, each joint controls the flow of fluid, pushing it through the lines in the right way because these channels should satisfy certain conditions.
  • Pipes as weights: In a neural network, the strength of a connection between two neurons is defined by weight. Therefore, it means that it influences the path of the signal throughout the network. In the same way, pipes can be considered weights with their dimensions, length, and diameter as well as material properties which determine resistance and thus, fluid flow velocity and pressure passing through pipe joints.

Potential Benefits and Challenges

Using a neural network methodology in managing pipe systems appears to be an interesting proposition for implementing a more intelligent and active control system of flow assurance. Considering the pipes as weights and joints as neurons, we might find it quite practical to put learning algorithms into practice that can self-adjust key system parameters through back-allocation on the fly, thus ensuring that it is most efficient, reliable, and safe under the given conditions, which is truly beneficial.

However, implementing this new strategy is not without its own set of challenges. One of the key drawbacks would be the loss of context that could hamper a better understanding of specific interactions and behaviours within the system, leading to difficulties in issue identification or a lack of basis for some decisions. Moreover, fluid dynamics, coupled with the physical limitations of pipes, further complicates matters. Building algorithms capable of correctly simulating and enhancing such a system would demand much study and experimentation: it stresses the importance that precision also has limits as an unwieldy countermeasure against highly dynamic complex systems.

Final Thoughts 

As we discussed, the potential for a groundbreaking transformation in the field of hydrodynamic process estimation within flow assurance lies in the adoption of the parallel calculation approach derived from neural networks. By leveraging the fundamental principles of neural network operation, particularly the mechanisms of forward calculation and adaptive learning, this approach presents a novel perspective on the optimization and management of fluid dynamics in piping systems. Nevertheless, the journey from theoretical concept to practical implementation is not without its obstacles and will surely need further experimentation and comprehensive investigation. 

To Top

Pin It on Pinterest

Share This