Deep learning-driven precision control of dilution rate in multi-pass laser cladding: experiment and simulation

The continuous energy input can lead to heat accumulation in the multi-pass lap laser cladding, which results in a progressive increase in the dilution rate and deteriorates the quality of laser cladding. Precisely controlling the stability of the dilution in the multi-pass laser cladding is still challenging. In this study, we proposed a deep-learning driven method for precisely controlling the dilution rate in the multi-pass laser cladding. Initially, the relationship between the dilution rate and power energy is retracted via the experiment-based finite element simulation. Subsequently, the convolution neural network deep learning is applied to optimize and improve the accuracy of the dilution rates in the cladding layer. The experiment verifies that the high stability of dilution rate in each pass, i.e., average errors of less than 10.88%, is achieved via in-situ adjusting of the power energy using the prediction obtained from the proposed method. We also attempted to provide insights into the dilution mechanism in Invar alloy multi-pass laser cladding as well as the potential applications of this method for other materials and other additive manufacturing.