Development of an optimization strategy for the design of structural parts manufactured by the QSP process “Quilted Stratum Process”
Topic(s) : Material and Structural Behavior - Simulation & Testing
Co-authors :
Juan ROJAS CARRILLO (FRANCE), François-Xavier IRISARRI (FRANCE), Cédric JULIEN , Denis ESPINASSOU (FRANCE)Abstract :
Composite materials allow to create lightweight structures used across several industry domains. Carbon fiber reinforced plastics (CFRP) are an example of high-performance composites commonly used in the aerospace industry. Though, their diffusion to other domains has been difficult by virtue of long processing times and high costs. To introduce a solution to these issues, CETIM developed the Quilted Stratum Process (QSP®). This is a multi-step fabrication process for high volume and production rate of thermoplastic composites. It is able to manufacture variable thickness and variable stiffness composite laminates by assembling composite plies of several orientations and shapes. The purpose of this work is to develop a design method for structural parts applied to QSP®. Consequently, the current work aims at optimizing a preform made up of welded patches, considering QSP® constraints. In this case, the optimization problem is combinatorial and implies both discrete and continuous, geometrical, and material design variables related to QSP®: the number and order of patches, the material and thickness, and the variables related to the shape and size of the patch. Moreover, other variables such as patches position, and material orientation. These many variables define an optimization problem with a complex research space. Hence, a multilevel framework was chosen to find the solution.
At the first step, a gradient-based optimization of the QSP® part is done to obtain an idealized target with optimal structural performance. This target is a homogenized stiffness (lamination parameters) and a thickness distribution. This step is followed by a discrete optimization step that targets the optimized thickness and stiffness distributions issued from step one (stiffness matching). This second step is based on the subdivision of the part into plies. The thickness distribution is sliced into a finite number of equal thickness plies. An Evolutionary Algorithm (EA) is used to find the best ply order and fiber orientation for each ply. Thus, the optimized solution at the end of the second step is composed of complete plies and also of plies that cover the surface partially. The obtained design can show an important stiffness distribution gap with respect to the target. To overcome this, we introduced a method to subdivide complete plies, based on mechanical considerations. This step, however, does not consider shape and manufacturability constraints, and might not guarantee compatible solutions with QSP®. Hence, a third step was added, aimed at simplifying shapes. These ply shapes define a new part zoning, compatible with QSP®. Next, the target design is updated (Step 4), using the new simplified zoning with a similar method as done in Step 1. The stiffness matching step is then repeated using the database of manufacturable ply shapes (Step 5) to obtain a final design whose mechanical behavior can be checked by FE analysis.
At the first step, a gradient-based optimization of the QSP® part is done to obtain an idealized target with optimal structural performance. This target is a homogenized stiffness (lamination parameters) and a thickness distribution. This step is followed by a discrete optimization step that targets the optimized thickness and stiffness distributions issued from step one (stiffness matching). This second step is based on the subdivision of the part into plies. The thickness distribution is sliced into a finite number of equal thickness plies. An Evolutionary Algorithm (EA) is used to find the best ply order and fiber orientation for each ply. Thus, the optimized solution at the end of the second step is composed of complete plies and also of plies that cover the surface partially. The obtained design can show an important stiffness distribution gap with respect to the target. To overcome this, we introduced a method to subdivide complete plies, based on mechanical considerations. This step, however, does not consider shape and manufacturability constraints, and might not guarantee compatible solutions with QSP®. Hence, a third step was added, aimed at simplifying shapes. These ply shapes define a new part zoning, compatible with QSP®. Next, the target design is updated (Step 4), using the new simplified zoning with a similar method as done in Step 1. The stiffness matching step is then repeated using the database of manufacturable ply shapes (Step 5) to obtain a final design whose mechanical behavior can be checked by FE analysis.