To succeed in the ultra-competitive LCD manufacturing industry, CSOT must deliver high-quality products in tight timeframes—but time-consuming product inspections were sapping its agility.
CSOT is bringing artificial intelligence (AI) to its factory, using IBM® Watson IoT™ technology to speed up the visual inspection process and detect product defects faster and more accurately.
Millisecondsto analyze products, shortening inspection lead-times
Cuts costsof production, helping drive more profitable operations
StrengthensCSOT’s reputation for product excellence
Business challenge story
Enhancing quality control
Whether it’s the latest smartphone or big-screen TV, global appetite for consumer electronics is booming. For manufacturers of LCD screens—key components in many of today’s electronic devices—the pressure is on to produce high-quality products to satisfy the demand.
For CSOT, a manufacturer of display components based in Shenzhen, China, staying ahead in such a fast-moving and competitive industry means continuously improving production processes and quality standards. The aim is to bring sophisticated, high-quality components to market ahead of its rivals, while driving down costs to protect profit margins.
To achieve these objectives, CSOT has been working hard to build smart factories, optimizing processes and harnessing the latest technologies to drive faster, more efficient operations. The company has succeeded in automating as much as 95 percent of the LCD manufacturing process. However, one bottleneck still remained: the all-important quality inspection stage.
A spokesperson elaborates: “Visual inspections are arguably the most crucial part of the entire manufacturing process. If we fail to spot defects before products are sent to device manufacturers, it could lead to costly product returns and rework, not to mention damage to our reputation for product excellence.
“Manual inspection methods are difficult to optimize and scale up. Our quality inspectors have to examine each LCD screen individually to check for flaws. This takes a considerable amount of time and, despite the fact that our inspectors are well-trained, there is still a risk of defects slipping through. At the same time, training people to become experienced inspectors takes significant time and resources.”
Bringing AI to the factory floor
To drive a smarter approach to quality control, CSOT introduced IBM Visual Insights—an AI-powered inspection solution that intelligently detects defects by comparing product images against a library of known defect images. Visual Insights is designed for easy integration with existing inspection processes, enabling CSOT to get up and running with the solution rapidly.
Working with an IBM research and development team, CSOT created a library of tens of thousands of pictures captured on its production line. The team classified the images into categories: good products and those with various kinds of defects. Next, they used Visual Insights to train an AI model that can discriminate between these categories.
CSOT applied this model to edge computing servers connected to ultra-high-definition cameras at an inspection point on the factory floor. As the cameras capture images of products at the inspection point, Visual Insights uses the AI model to rapidly compare them against the appropriate defect images and classify them accordingly. The result of the classification is then sent to the cloud for review and evaluation by a human inspector.
Visual Insights provides a confidence level for every image it classifies, ranging from zero (no match) to 100 percent (a perfect match). If the confidence level is below an acceptable threshold, the system prompts an inspector to review the item and determine whether there is indeed a defect. This capability helps reduce inspection times and costs by allowing CSOT to focus human expertise only on items that truly require it, while relying on intelligent visual recognition for the majority of cases.
As an AI solution, Visual Insights is learning all the time. It continuously takes feedback from the inspection team, who use their years of expertise to review and evaluate its automated classifications. The corrective information, along with the image from the factory floor is then included in the next training cycle for the AI model, improving its ability to detect future defects.
Boosting production speed and quality
By combining AI technology with human expertise, CSOT is driving more accurate product inspections, which helps to minimize the risk of potentially defective products leaving the production line, boosting overall product quality. This will reduce costs, improve manufacturing throughput and enable the company to maintain high quality standards, protecting its reputation for product excellence.
In addition, smart inspection capabilities have enabled CSOT to accelerate what was a previously tedious and time-consuming manual task. Visual Insights can analyze product images in milliseconds—thousands of times faster than a human operator. This helps CSOT to identify defects with speed and confidence, shortening inspection lead times.
A spokesperson concludes: “At CSOT, our priority is to use innovative technologies to provide consumers with the best quality products. IBM Visual Insights helps us take operational excellence to the next level. We look forward to continuing our cooperation with IBM, and to using Visual Insights to fully enable intelligent manufacturing.”
About Shenzhen China Star Optoelectronics Technology Co., Ltd.
Shenzhen China Star Optoelectronics Technology Co., Ltd. (CSOT, also called Shenzhen Huaxing Photoelectric Technology) is one of the world’s leading designers and manufacturers of display panels. Founded in 2009 and based in Shenzhen, China, CSOT operates as a subsidiary of TCL Corporation—one of the world’s biggest consumer electronics producers.
Take the next step
To learn more about IBM Visual Insights, please contact your IBM representative or IBM Business Partner, or visit the following website: ibm.com/marketplace/visual-inspection-for-quality
IBM is an established leader in the Internet of Things with more than 6,000 client engagements in 170 countries, a growing ecosystem of over 1,400 partners and more 750 IoT patents which together help to draw actionable insight from billions of connected devices, sensors and systems around the world. For more information on IBM Watson IoT, please visit ibm.com/iot