Model Architecture

  • Bayesian 3D CNN architecture

Enables estimation of real-valued process parameters for Root Cause (RC) isolation

_images/3d_cnn.png

Fig 1: Bayesian 3D CNN for Object Shape Error Response (OSER) for single-station assembly systems

  • Bayesian 3D U-Net architecture

Enables estimation of heterogeneous set of real-valued and categorical process parameters and estimates upstream shape error for Root Cause (RC) isolation.

_images/U_net.png

Fig 1: Bayesian 3D U-Net for Object Shape Error Response for multi-station assembly systems (OSER-MAS)