Identification method of abnormal welding point of flat induction coil

Abstract

The invention discloses a method for identifying abnormal welding points of a plane induction coil, comprising the following steps: using etching and other technologies to make a plane induction coil, the finished product is transmitted according to industrialized production steps; when transmitted to a surface defect detection device, The camera shoots the product and collects high-definition images of the product surface; the high-definition images are transmitted to the abnormal welding point extraction equipment through the internal bus; the equipment is compared with the abnormal welding point feature library, and a confidence value with abnormal characteristics is given; the confidence value is passed. The internal bus is passed to the industrial robot arm and above a certain threshold it is diagnosed as a flat induction coil with abnormal solder joints. The invention can effectively integrate the traditional image classification model structure, improve the recognition efficiency of abnormal welding points of the plane induction coil, can effectively solve the serious class imbalance problem in the detection of defective products, and has strong practicability and robustness.

Description

technical field [0001] The invention belongs to the field of product surface detection in industrial manufacturing, and particularly relates to a method for identifying abnormal welding points of a plane induction coil. Background technique [0002] Planar electromagnetic coil is an important part of various micro-motors, micro-electromagnetic sensors and actuators in micro-electromechanical systems. The production process of planar electromagnetic coils mostly adopts LIGA and other processes, and the LIGA process requires the use of a synchrotron radiation light source and an X-ray mask. The planar electromagnetic coil produced by the etching process has problems such as coil virtual welding and broken welding. The detection of defective products can be carried out by various means such as manual visual inspection. However, the traditional method has problems such as high inspection cost. [0003] Deep learning-based surface detection methods can learn defect depth featur...

AI Technical Summary

Problems solved by technology

At present, there are many deep learning feature learning schemes. However, most of these schemes are aimed at classifying images with balanced classes, and have not conducted in-depth research on actual industrial problems with severe class imbalance.

Even though there are deep learning loss functions that have been studied for the class imbalance problem, there is no specific research done on the surface defect detection problem to be done

Images