SYSTEM AND METHOD FOR POWER MODULE DEFECT DETECTION
In an embodiment, a method includes: capturing a first image of a power module, the power module including a power electronics circuit, the power electronics circuit including power semiconductor dies; identifying positions of the power semiconductor dies in the first image with a die detection model; extracting second images of the power semiconductor dies from the first image according to the positions of the power semiconductor dies in the first image; and identifying defects of the power semiconductor dies in the second images with a defect detection model, the defect detection model being different from the die detection model.
1 . A method comprising:
training a defect detection model with first images of first power semiconductor dies, the defect detection model comprising a convolutional neural network;
forming a power module comprising second power semiconductor dies;
capturing second images of the second power semiconductor dies; and
identifying defects of the second power semiconductor dies in the second images with the defect detection model by processing the second images with the convolutional neural network to simultaneously predict bounding boxes of the defects in the second images and predict class probabilities for the bounding boxes.
2 . The method of claim 1 , wherein the power module further comprises a package substrate, and the second power semiconductor dies are mounted to the package substrate.
3 . The method of claim 1 , wherein the second power semiconductor dies are silicon carbide dies.
4 . The method of claim 1 , wherein capturing the second images comprises sensing ultraviolet light rays with an optical microscope.
5 . The method of claim 4 , wherein the power module further comprises a passivation layer on the second power semiconductor dies, the passivation layer being transparent to the ultraviolet light rays.
6 . The method of claim 1 , wherein the defects comprise cracks in the second power semiconductor dies.
7 . The method of claim 1 , further comprising storing the defect detection model in a memory of a controller.
8 . A system comprising:
a processing tool configured to form a power module comprising power semiconductor dies;
a camera configured to capture images of the power semiconductor dies; and
a controller comprising a memory storing a defect detection model, the defect detection model comprising a convolutional neural network trained with first images of first power semiconductor dies, wherein the controller is configured to:
receive second images of second power semiconductor dies of the power module from the camera; and
identify defects of the second power semiconductor dies in the second images with the defect detection model by processing the second images with the convolutional neural network to simultaneously predict bounding boxes of the defects in the second images and predict class probabilities for the bounding boxes.
9 . The system of claim 8 , wherein the power module further comprises a package substrate, and the power semiconductor dies are mounted to the package substrate.
10 . The system of claim 8 , wherein the power semiconductor dies are silicon carbide dies.
11 . The system of claim 8 , wherein the camera comprises an optical microscope configured to sense ultraviolet light rays.
12 . The system of claim 11 , wherein the power module further comprises a passivation layer on the power semiconductor dies, the passivation layer being transparent to the ultraviolet light rays.
13 . The system of claim 8 , wherein the defects comprise cracks in the power semiconductor dies.
14 . The system of claim 8 , wherein the controller is further configured to control the processing tool to stop a manufacturing process in response to identifying the defects.
15 . A controller comprising a memory storing a defect detection model, the defect detection model comprising a convolutional neural network trained with first images of first power semiconductor dies captured by a camera, wherein the controller is configured to:
receive second images of second power semiconductor dies of a power module from the camera; and
identify defects of the second power semiconductor dies in the second images with the defect detection model by processing the second images with the convolutional neural network to simultaneously predict bounding boxes of the defects in the second images and predict class probabilities for the bounding boxes.
16 . The controller of claim 15 , wherein the first power semiconductor dies and the second power semiconductor dies are silicon carbide dies.
17 . The controller of claim 15 , wherein the controller is configured to identify the defects by:
processing the second images with the convolutional neural network to output tensors, each tensor including coordinates of a bounding box and a confidence value; and
using tensors with confidence values greater than or equal to a threshold.
18 . The controller of claim 15 , wherein the power module further comprises a passivation layer on the second power semiconductor dies, the passivation layer being transparent to ultraviolet light rays.
19 . The controller of claim 15 , wherein the controller is further configured to control a processing tool that forms the power module to stop a manufacturing process in response to identifying the defects.
20 . The controller of claim 15 , wherein the defects comprise cracks in the second power semiconductor dies.