The use of AI for quality monitoring and control is on track to generate US$418.2 million in sales by 2025, with Asia Pacific (US$133.8 million) and North America (US$117.1 million) expected to see the highest growth, according to a report by market intelligence firm Tractica.
The use of AI is still in a relatively nascent stage, but
manufacturers have picked up the pace in the implementation. A key driver is the pursuit of making quality products.
The use of AI for quality monitoring and control is on track to generate US$418.2 million in sales by 2025, with Asia Pacific (US$133.8 million) and North America (US117.1 million) expected to see the highest growth, according to a report by market intelligence firm Tractica.
Nowadays manufacturers ensure quality by establishing certain parameters and final product tolerance, so that finished products come as close as possible to the original design and performance specifications.
Often times manufacturers want to ensure their equipment is operating properly and that no variabilities in materials or processes are introduced to affect the products made.
In factories,
with sensors attached to the machines, signals can be gathered and monitored to keep track of the machines. Ambient sensors will measure temperature, humidity and other environmental data. When conditions affecting quality are detected, they can be corrected by manufacturers before faulty products are made.
One real-world example comes from an automaker that completes more than 50 million welds per day. Signals are collected off
welding robots and the company will know if the quality of the weld is good or bad.
Signals in known good welds are compared to signals from the robots in real time. With the help of machine learning, when new anomalies are detected, they are fed back into the system to refine and improve the algorithm. The end result is a 98% detection rate for bad welds, saving much cost and time, Tractica says.
How product inspection is done
When it comes to quality control, typically visual inspections were performed by the manufacturing staff. However, humans can suffer from fatigue and degrading health conditions, which may compromise their performance.
Machine vision and
machine learning can come into play, improving the efficiency and accuracy of the inspection process.
Landing AI, founded by Andrew Ng, a machine learning expert who led the development of Google Brain, is company that incorporates machine vision, object identification and machine learning to help manufacturers integrate AI into their workflows. Landing AI’s solution can inspect a part in half a second and recognize patterns of imperfections, often more accurate than humans, according to the company.
The company can reportedly recognize patterns of imperfections after reviewing just five product images, compared with traditional visual inspection systems that must be trained with around one million images.
Fujitsu Laboratories is another case in point. When changing a manufacturing line or part, often the inspection system must also be revised. The problem is that the amount of image data that was usable prior to starting new manufacturing lines was limited.
Fujitsu turned to specialized genetic programming to speed up the learning process by configuring the image recognition application with templates, Tractica says. “Templates can narrow the learning and recognition process to three processes: image enhancement, threshold process, and binary image handling. The program will learn and evolve automatically by preparing training data from images of normal and defective parts to make pass/fail judgments.”
“When tested on a parts assembly line, the new image recognition system automatically generated code for inspection and achieved a nearly 100% recognition rate,” Tractica says. As a result, the time to develop programs for new inspection processes decreased by 80%.