Kodak Implements Neural Inspection

Eastman Kodak (Rochester, N.Y.), a manufacturer of imaging-related products, has developed an online neural network-based machine vision system for surface mount solder paste inspection. This Windows-based inspection system automatically inspects and analyzes the fine-pitch solder paste physical quality.

The system, developed by Kodak's automatic machine systems division, uses digital camera data in conjunction with a neural network to grade the solder paste application process. Neural networks are intelligent software that are able to adapt over time to improve their own performance. They are particular useful at identifying relationships in data.

Solder paste is placed on a circuit board to affix surface mount components. A subsequent manufacturing step places the component on top of the surface mount solder paste, which then passes through a reflow oven bonding the part to the board.

A neural network was trained using the solder paste image data. In addition, a neural network trained on parametric data extracted from the images was also used to grade the physical quality of the solder paste. Analysis of both the solder paste images and parametric data, as graded by the neural network, showed very good correlation with a human expert.

The overall online system displays trend graphs of both the neural network quality score, in addition to the various parametric data items extracted from the images, such as pad width and area. The trend display of the solder paste quality will be used for both statistical process control and basic engineering understanding of the solder paste operation for use in the continuing improvement cycle.

Click here to return to Table of Contents for the Intelligent Manufacturing April issue.