October, 2023
www.us -
tech.com
Inspection Using AI: New Approaches in Quality Control
Continued from previous page
enced operators must correctly assess and clas- sify the borderline cases sorted out in the auto- mated process. The difference between “prod- uct already defective” or “product just accept- able” can be minimal, which is why collected “image bases” with labeled verification results provide valuable assistance. Feeding AI models with this information
opens up entirely new possibilities for effi- cient workflows thanks to deep learning. In a first step, the AI can be integrated as a sec- ond opinion. This means that the software works inconspicuously in the background in a “dual control principle” mode to check the operator’s classifica- tion. If the artificial intelligence notices a deviation from its own decision, it outputs a confidence level in percent, indicating how certain it is of its assessment of the individual case at hand. However, the final decision for the classification still lies with the operator. Human error — e.g., due to
fatigue — is prevented in this way and the quality of the products and amassed data is further improved. All classifications of the AI and the operator are stored and can be retrieved later for evalua- tions. This serves very well to com- pare the operator’s skills and abil- ity with the AI’s performance. If the AI proves to be particularly reliable in partial areas, it can be given decision-making power in this clearly defined field — and limited only by involving a super- visor, for example. In further steps, a white list defines where the AI is allowed to act completely autonomously. Viscom uses this kind of flexible AI-supported veri- fication with its vVerify classifica- tion software.
Targeted Segmentation A major advantage of AI in
image processing is that it is not distracted by interfering factors and recognizes objects with high precision. These can be the paths of wire bonds or the anodes and cathodes of battery cells. In this context, the term segmentation is heard very often. Basically, it is a matter of
deciding in a smart automated process which pixel still belongs to a certain object and which does not with a very high degree of certainty. The information is then available to other inspection algorithms, which continue to work with it accordingly. In the field of X-rays, for
example, voids in solder joints can be segmented and thus precisely identified and measured. Distur - bing shadows in the X-ray image are accurately retouched by the application. Not so long ago, slice images from 3D X-ray volume information have become popular and established in practice, but depending on the specified inspec-
AI retouched image, clearing away the shadow of a bracket covering a BGA.
Page 61
tion criteria, AI now unlocks opportunities to compete with them in 2D and with faster cycle times. Both approaches can also be com- bined if required. Artificial intelligence is also very good
at recognizing component shapes. This is done on the basis of collected information about forms and geometries and enables even easier creation and optimization of inspection programs. The correct inspection patterns are also
assigned automatically. If the appropriate tools and information are readily at hand, an inspection program for automatic optical inspection can be created in just a few min-
Continued on page 68
CONNECTING THE FUTURE OF SOLDERING CONNECT SMART
SAFE
www.weller-tools.com/wxsmart
Hall A4, Stand 241 November 14–17, 2023
Visit us
www.weller-tools.com/wxsmart See at productronica, Hall A4 Booth 241
CONTROL
Apex Tool Group, LLC 1000 Lufkin Road Apex, NC 27539 Tel: +1 919 387 0099 Fax: +1 866 907 9706 ©2023, Apex Tool Group LLC
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88 |
Page 89 |
Page 90 |
Page 91 |
Page 92 |
Page 93 |
Page 94 |
Page 95 |
Page 96 |
Page 97 |
Page 98 |
Page 99 |
Page 100
