Shaped Steel Catenary Riser Design Optimization Using Artificial Intelligence

Abstract SSCR (Shaped Steel Catenary Riser) is achieved by the targeted application of buoyancy to an SCR. SSCR configuration provides better resistance to compressive loads, has improved strength and fatigue responses compared to SCR configuration and has reduced CAPEX costs compared to a generic SLWR configuration. It also reduces RFTP (Riser to Flowline Transition Point) loads compared to a SCR and thereby, eliminating the need for seabed anchoring completely or making it simpler; e.g. with a mudmat type solution with pin-piles. In contrast, due to the large loads imparted by SCRs, expensive solutions like tethering to suction piles are typically required. SSCR solution aids in faster installation with an easier pre-lay solution and is also scalable for water depth and pipe size providing flexibility to the projects. However, optimizing the shape of the SSCR requires significant number of iterations to satisfy constraints related to strength, fatigue and other competing requirements, making it a complex task. Therefore, this paper presents the application of Particle Swarm Optimization Algorithm (PSO), which is a bio-inspired technique to enhance the ability to find an optimum ‘S’ shape for the SSCR solution. SSCR is studied for different water depths and different sets of PSO using Gulf of Mexico Metocean data. The solution space for an optimal configuration is defined in terms of the length to start of buoyancy section and length of buoyancy section. The objective function in the algorithm depends on the SSCR’s strength response, fatigue response and RFTP loads. The configuration solutions with poor interference result, low seabed clearance and unrealistic elevation between sag and hog regions are the constraints, which are penalized in the algorithm and rejected. The PSO algorithm is implemented in Python. The constraint parameters, strength responses, and fatigue responses are passed from OrcaFlex to Python to calculate the fitness of each particle position. The PSO algorithm then calculates the next iteration of particle positions and velocities based on the personal best position of each particle and global best position of the population. This process is repeated for several iterations to obtain the optimal SSCR configuration. The PSO algorithm finds an optimal configuration within the given solution space and meets all the constraints including von Mises stress, standard deviation of tensile stress, minimum seabed clearance and allowance clearance in the clash check. Furthermore, different sets of the swarm particles have found an optimal configuration with nearly the same fitness value and minimal difference in the results indicating an ideal quantity of the particles to reduce the computational time. SSCRs generated using PSO can be used as a starting point for the project studies and helps engineers to come up with an initial design capturing the basic requirements. It has the potential for improved decision-making, meeting fast-track project demands, streamlining the design and providing CAPEX benefits. A database of optimal SSCR solutions generated using PSO can be used to train an artificial neural network to generate optimized SSCR configurations. Once trained and benchmarked, the trained AI system can provide efficient SSCR solutions depending on the project requirements and constraints.