Ting Lan, the General Manager of Unimicron, revealed that the company’s product inspection efficiency has increased by 70% since the introduction of its new AI flaw detection system.

 

Ting Lan said that an AVI automated optical inspection machine was formally used by Unimicron to detect defective products. The inspection results were often affected by product color variances, differences in height and other factors, forcing the company to deploy an additional 11-15% of their workforce to recheck the results obtained by the automated optical inspection machine. However, manual testing tends to be subject to too many variables caused by “personnel fatigue.” A detection error could cause the company to lose hundreds of millions of dollars.

 

According to Unimicron, the company generates millions of data records, from product development to production. Unimicron has long contemplated about how to apply this data to optimize production processes and has not stopped watching the development of market technology.

 

Recently, the company has decided to follow the trend by introducing AI due to the maturity of its relevant applications and tools. It has combined deep learning and an AVI machine to perform product testing on its production line. Ting Lan noted that the introduction of AI has helped the company to deploy only 30% of the workforce while conducting the manual inspection, meaning that “efficiency has increased by 70%.” However, he is still not satisfied. “Unimicron’s ultimate goal is unmanned detection.” Unimicron has also started conducting data collection and analysis, hoping that this can help the company discover which process have gone wrong by comparing data variations and production results.

 

What is the next move in the application of AI? Ting Lan pointed out that M2M (machine to machine) is desired by the manufacturing industry to generate automated links through “machine-to-machine dialogue.” For example, when a machine detects a problem with a product, another machine can handle it automatically.

 

Another thing Ting Lan wants to achieve with AI is security detection and prediction, such as predicting whether an operator’s actions could violate the regulations, or monitor for environmental changes in the plant environment such as temperature, odor, noises, etc. to avoid a malfunction or other dangerous occurrence.

 

However, Ting Lan noted that not only does the industry lack AI talents but it also has difficulty in implementing talent management after the “realization of AI.” After the introduction of AI, the manufacturing industry in particular will face many new changes, such as changes to production and management processes. Therefore, it is necessary to understand “talent planning” of human-computer collaboration to monitor and manage AI and its processes (news source: iThome)