Layer time optimization in large scale additive manufacturing via a reduced physics-based model

In large-scale additive manufacturing (AM), ensuring product quality and production efficiency has been dependent on the skills and experiences of machine operators, and there has been a lack of guidelines based on accurate data and a model from systematic analyses. The product quality and the production efficiency are highly influenced by layer deposition time (a.k.a. layer time). The determination of a proper layer time involving a high-fidelity model requires high computational cost, and cannot be utilized for an online feedback system where fast temperature prediction is necessary. In this work, we propose a fast layer time optimization framework utilizing a reduced physics-based one-dimensional heat transfer model to predict the cooling behavior and layer temperature. We also perform a high-fidelity three-dimensional finite element analysis (FEA) with two geometries involving large angles and sharp angles. The temperature from the reduced model is adjusted by variances calibrated based on the FEA model reflecting geometric effect so that the prediction from the reduced model can be applied to complex geometric designs. This process of temperature prediction is named the hybrid model, and it allows the offline design of layer time optimization. We combine the temperature data into an optimization model, which monitors the temperature of multiple positions and balances the relationship between the layer time and the layer temperature. We also develop an iteration -based solution approach by combining the layer time optimization model with the hybrid model. The approach involves iterations between the proposed layer time from the optimization model and the temperature predicted from the hybrid model until the predicted temperature converges to a target layer temperature, determining an optimal layer time. We apply the developed process to two cases with different printing geometries: hexagon and star shapes. This paper provides a simplified and lower-cost methodology to determine an optimal layer time and improve product quality in the large-scale AM process.