Speaker
Description
The design of superconducting magnets demands a multi-phase, multi-scale and multi-physics approach which can be more or less challenging according to the complexity of the requirements. According to the different schools, the design effort is usually performed in three or four phases: we can identify them starting from the simplest to the most complex as feasibility design, conceptual design, preliminary design, detailed design, with feedback loops in each phase. If we focus on the design of the cryostat and of the cold mass, the simplest loop consists in six steps: at first the design efforts focus on the magnetic design (step 1), which determines the parameters for the conductor design (step 2). Once the conductor and the geometric shape of the magnet are defined, the quench protection and transient analysis follow (step 3). Finally, the mechanical design of the cold mass (step 4) and the mechanical (step 5) and cryogenics design (step 6) of the cryostat complete the loop, with possible variations of the order of the steps according to the challenges the design encounter.
Compromises between each step are needed to guarantee both the compatibility of the design respect to the requirements and its physical and engineering feasibility.
The main limits of this approach are that: a) it is time consuming; b) according to the feasibility of the manufacturing (i.e., fabrication tolerances, conductor performances, etc.), sometimes the preliminary or the detailed design need to be modified. CEA is currently working on optimizing the design process of superconducting magnets with a holistic approach: during the feasibility and conceptual design phases we largely explore the parametric space of the design using the six steep loop and we guide the design with artificial intelligence algorithms. The principle behind is not only to find multiple local minima where the requirements are satisfied, but also to look at the curvature of the parametric space around these minima: the milder the curvature, the larger the tolerances that can be considered without changing the nominal design.
In this paper, the CEA approach is presented with an application example on the Spin Rotators Solenoids for the Electron – Ion Collider (EIC), where the conceptual design has been achieved thanks to ALESIA, the in-house artificial intelligence tool developed to guide the design process of superconducting magnet.
