Alcoa Automates a Production Process with a Solution from SmartSoft

Overview

Alcoa is a world leader in the production of primary aluminum and fabricated aluminum and also the largest miner of bauxite and alumina refiner. For over 20 years, the company has facilitated many improvements in various areas, such as automotive, aerospace, industrial, consumer electronics, packaging, commercial transportation, and construction. Alcoa currently employs around 61,000 people in more than 200 locations in 32 countries.

The company needed to automate one of their production processes in the Howmet-Morristown (TN) division. SmartSoft was selected to implement an OCR (optical character recognition) solution to aid in solving the problem.

Case Problems

Alcoa Howmet-Morristown supplies complex ceramic cores forming the internal cooling passages of turbine airfoils.

The constructors were scanning a part with a 3-D scanner and the serial number on the part had to be automatically extracted. They used a scanner to create images of a 3-D stamp on a flat surface with a mortar-type material that was poured around it. The Alcoa professionals were looking for optical character recognition software that would recognize the images and turn them into actual text, enabling them to fully automate the process.

   

Solutions

SmartSoft was selected to implement a robust solution based on the Smart OCR technology to help Alcoa automate the process. The solution’s watched folder feature was used to aid the automation – the scanned images were sent to the watched folder, and automatically queued for recognition.

To achieve optimum recognition results, the developers extended the OCR engine by adding a custom glyph set and limited the number of symbols in the software database to that of the ones used on the stamps, so that fewer symbols were available for comparison.

A custom noise filter based on Connected Component Analyses was added to ensure that the symbols were recognized properly. Since some of the scanned images were slanted, the developers also created a custom auto rotation algorithm that worked with single words. A custom algorithm for splitting groups of more than three glyphs was created to fix the issue of connected symbols.

Since the scanned images were different sizes, dynamic scaling was initially applied to automatically detect the size of the symbols. However, on some of the parts, the mortar created bubbles which appeared as small black dots on the images and the OCR software could recognize these bubbles as the number zero. To solve this problem, the size of each symbol was automatically compared to that of the other symbols and not counted as a number unless it was at least 50% of their size. This ensured accurate results for any size of the scanned image. Finally, Alcoa opted the dynamic scaling out and decided to go for set image sizes – since they felt it was easy and fun to work with SmartSoft on tweaking the product at any time they felt this was necessary.

Results

The initial goal was achieved – Alcoa was able to successfully automate the process of scanning and recognizing the 3-D stamps. Optimum recognition results were ensured as a result of the collaboration between the two companies. Daryl Lewis of Alcoa commented: “Thanks for the support! You guys were very helpful and we are definitely recommending you to any of our other locations that may have a need for this type of software!”

More information on Alcoa is available at alcoa.com.