In this study, we focused on estimating the penetration resistance of closed end pile by deep learning using visualized images of penetration resistance with force chain and particle motion during pile penetration.

The learning data of the visualized images were collected through the process of pile penetration into the mechanoluminescent-coated particle assemblages that emitting the force-induced luminance. The wedge-shaped higher luminance intensity area were obtained under the pile shown in left side figure.

We developed deep learning model of following two types: generating the estimation image of force distribution during pile penetration based on the learning of the visualized images, estimate penetration resistance based on the force visualized images. These estimated results was verified through the captured images and measured value using load cell.

The result of verification for generated images shows the higher luminance intensity area under the pile end. The result of verification for estimating penetration resistance was shown in right side figure. The estimated value shows good agreement with measured value, whose error was around 10%.