One day in 2019, an executive from a top mobile phone manufacturer led a group to visit AUO Optoelectronics to exchange and found that when the company held a morning meeting, all employees of this company held tablets and reported with data reports. Paper and pen were gone in the morning meeting of this company, which shocked the executive.
As the world's top mobile phone brand and manufacturer, this manufacturer built its own mobile phone production factory very early and began to explore in the field of smart manufacturing early. At that time, they had run multiple digital systems in one of their production workshops.
However, the introduction of digital technology has not completely changed the previous production management model of this mobile phone manufacturer. In the usual meetings, they still used "old" production tools such as paper and pen to report and conference records. The digital systems that have been introduced in large quantities have not been opened up at the operation level and cannot be linked. The data recorded by the digital system is ultimately only stuck in the system. The situation of this mobile phone manufacturer is not an exception in China. In the past few years, domestic manufacturing companies have introduced a large number of digital technologies in operation, management, production and manufacturing. After these digital technologies have been introduced into the company, they have not made the company look brand new as everyone expected.
In fact, when traditional manufacturing companies realize the importance of digitalization and start introducing technical optimization processes, the real difficulties have just begun.
Data trough build "high-rise buildings"
In 2015, China's manufacturing industry pressed the fast-forward key of digital transformation , and the epic of traditional manufacturing was wrinkled by the wind.
AUO Optoelectronics, which focuses on the field of display technology, is one of the first companies to feel the potential disruptiveness of digital technology that is coming. Therefore, this year, AUO Optoelectronics built an intelligent manufacturing demonstration workshop in the factories of Suzhou , Kunshan , and Xiamen , and began to try to transform to the new model of intelligent manufacturing internally.
In the following two years, AUO Optoelectronics gradually connected the old equipment to the Internet and introduced a digital system to collect data, analyze data, and optimize the process. However, this process is not easy. After AUO Optoelectronics initially transmitted data to the service layer, the data utilization rate has been very low, basically maintaining it at around 5%. How to effectively utilize these data in so that data can play the role of data assets in the manufacturing operation management efficiency has become an urgent problem for AUO Optoelectronics.
Zhao Lina, who was the intelligent manufacturing director of AUO Optoelectronics at the time, quickly realized this problem, so she made the first three-year plan for AUO Optoelectronics that was transforming at that time - using data to drive business transformation.
In the following three years (2015-2017), AUO Optoelectronics began to sort out the data of each business unit with "data-driven" as the core, and began to promote the interconnection of all business data, so as to transform to digitalization.
For any enterprise, digital transformation is a grand proposition of the times that will be faced sooner or later, especially for traditional manufacturing, when it involves specific businesses, this issue becomes more messy and complex.
Zhao Lina told 一王, "We considered many solutions at that time, such as how to open up the warehouse data to build a smart warehouse, how to open up the production line logistics data to form intelligent logistics, how to open up the quality data to achieve smart product protection ... At that time, we introduced the Internet of Things technology and the artificial intelligence platform, internally promoting the transformation of data from simple analysis presentation to complex prediction and pre-control."
All technological revolution is essentially a talent revolution.
If the traditional manufacturing industry wants to truly get out of the old model and replace it with a new power engine, it is necessary to completely reconstruct the knowledge system of everyone in the original team. Zhao Lina's understanding at the time was that " allows everyone in the business team to transform into a data analyst ".
Therefore, when AUO Optoelectronics began digital transformation in 2015, an internal training system for business teams was gradually established. From the data-driven business principles to the application of data analysis tools, from the construction process of each business plan to the review of the final implementation plan, in the past three years, the popularity of data analysis-related technologies in AUO Optoelectronics' business team has been continuously increasing, and most members of the business team have the level comparable to professional data analysts. The transformation of the
business team has made AUO Optoelectronics' digital transformation no longer superficial. Since then, in management meetings of all sizes, employees no longer use paper and pen when conducting work reports, but naturally transitioned to the "digital model" of tablets + data reports.
This change from production tools to team thinking mode was one of the main reasons why the top mobile phone manufacturer later made a decision to purchase their consulting services without hesitation after visiting AUO Optoelectronics. Based on this, AUO Optoelectronics later proposed a "5U" methodology for the digital transformation of business personnel.
The arrival of the new production model seems to have not experienced any magnificence and drastic efforts. All the hardships and pains are as quiet as the deep sea torrent.
can show us more intuitively "one result" and "one data" - when the first three-year plan was completed, AUO Optoelectronics has connected the data of all levels and functions and built a complete Industrial Internet . At the same time, AUO Optoelectronics' equipment networking rate has increased to 87%, labor cost savings have been 42%, and product yield has been increased by 2%.
After the first three-year plan was over, AUO Optoelectronics also came up with another idea to "transform results" the transformation experience and methodology accumulated over the past three years to serve more companies and even the entire industry.
So, this year, AUO Digital began to build, and the story of the service transformation of an optoelectronic enterprise has begun.
The traditional production model subverted
In October 2018, AUD Digital (formerly named: AUD Intelligence) was officially established as a wholly-owned subsidiary of AUD Optoelectronics that provides intelligent manufacturing consulting services and complete solutions to the outside world, and Zhao Lina became the general manager.
"We and AUO Optoelectronics use a set of methodology and technical frameworks. This methodology is not static, but will be updated every three years. The new methodology will only provide corresponding digital solutions to enterprises after AUO Optoelectronics is verified and developed and matured," Zhao Lina told Zhiding.com.
In fact, just in the year AUO Digital was established, AUO Optoelectronics' second three-year plan for intelligent manufacturing was officially launched, and the theme of the second three-year plan was positioned as "AI-driven".
If big data brings a change in thinking to the manufacturing industry, artificial intelligence brings a real subversion of production models.
As a well-established LCD panel manufacturer, AOO began to carry out information construction very early. As early as 2002, it used a series of software systems such as ERP (enterprise resource planning system), MES (manufacturing execution system), EMS (equipment management) and other software systems. However, due to the immature development of artificial intelligence technology in the early years, there was no way to talk about AI quality inspection. AOI was still the mainstream method of replacing artificial quality inspection in the field of surface defect detection at that time.
AOI, that is, automatic optical detection, and the similarity to AI quality inspection is that it also requires visual algorithm drivers - taking photos of each product on the production line through the camera, and then comparing the photos of each product with the product pictures without defects in the standard sample library built in advance to see if there are any differences. Defect detection methods such as
naturally have one disadvantage - the quality of the visual algorithm directly determines the detection rate, which means that if the AOI algorithm only has a detection rate of 50% in the actual production environment, the remaining 50% of the products still need to be manually re-judged.
As one of the three largest LCD panel manufacturers in the world, although AUO Optoelectronics' production equipment automation level is already very high, it still requires a lot of labor costs to conduct manual re-judgment for the detection of LCD panels used in consumer electronics. This cost once accounted for more than one-third of AUO Optoelectronics' total labor costs.
Because of this, AOI Optoelectronics developed its own AOI detection machine in its early years. By introducing AOI detection methods, AOI Optoelectronics saved 70%-80% of its labor costs in the quality inspection process. However, such technological upgrades also bring unexpected "engineering costs".
"Because the AOI detection method cannot guarantee the accuracy of the detection every moment, in order to avoid our defective products from flowing into the client, we will even raise the accuracy of the analysis and judgment to a relatively high level, which requires the introduction of a large number of engineers to process the image."
With the production volume of AOI Optoelectronics, the number of pictures that need to be processed every day was basically maintained at the tens of millions, and the "engineer cost" has thus become a new bottleneck for manufacturing defect detection to reduce costs and increase efficiency. The breakthrough in the bottleneck of
was after the emergence of AI quality inspection.
In 2017, after several years of rapid development of artificial intelligence technology, it began to enter the industrial dividend period. The most prominent thing in the manufacturing industry is the disruptive changes brought by AI quality inspection to the quality inspection industry.
Relatively speaking, the essence of in the past was to "find different" and AI quality inspection was exactly the opposite, and the essence was to "find similarity".
AI quality inspection is to learn standard samples and defect samples through deep neural networks, and form judgment standards based on the similarity of the picture, thereby automatically determining whether the products on the machine have defects. Based on this working principle, the algorithm model of AI quality inspection will continuously improve the model accuracy through self-learning during use. The emergence of the self-evolution ability of
AI algorithm has begun to subvert the unchanging traditional production model.
In fact, when AUD was established, AUD's technical team and Baidu 's cooperation in the field of AI quality inspection has lasted for a year, and AUD Optoelectronics has also begun to introduce AI quality inspection within. In December 2018, when standing on the podium of the 2018 Baidu Cloud Intelligence Summit, Zhao Lina pointed out that " has been introduced to the public. Through the introduction of AI quality inspection, the misjudgment rate of AUO Optoelectronics' entire quality inspection process has been reduced by 50%, and the retrial and technical maintenance personnel have also been reduced by 50%. ."
The cooperation between AUO and Baidu is just a historical epitome of the massive entry into the manufacturing industry. In this year, major Internet giants began to expand their industrial territory to the virgin land of manufacturing based on cloud business.
IT manufacturers can't handle the "scenario"
At the end of 2017, Ali released the industrial Internet cloud platform and designated Guangzhou as Alibaba's national industrial cloud headquarters. The following year, the next year, supET and Feixiang were released in Zhejiang and Chongqing. The land grabbing of Alibaba's industrial Internet cloud platform has become a true portrayal of their entry into the manufacturing industry.
Compared with Alibaba's long-term development, Baidu is more cautious in its manufacturing industry. It is more about taking advantage of the dividends of artificial intelligence. After establishing cooperation with star companies in various manufacturing sub-sectors such as AUO Optoelectronics and Baowu Steel, it first used AI quality inspection as a breakthrough point and released the quality inspection cloud in May 2018.
It is not just domestic Internet giants such as BATH that are pouring into the manufacturing industry. Foreign countries, tech giants such as Microsoft , Google , Amazon are not idle either, and have entered the traditional manufacturing industry based on their original cloud computing business.
As a personal practitioner of intelligent manufacturing, Zhao Lina has witnessed the ups and downs of this industry and has also conducted in-depth cooperation with Internet giants such as Baidu, Microsoft, and Amazon. She is well aware of the disruptiveness of artificial intelligence to the manufacturing industry and the vitality brought by technology giants to the manufacturing industry, and also understands the gap between the two industries.
"IT enterprises are good at underlying technology, and their role is more in line with their role, which is actually the enabler of underlying technology," Zhao Lina told Zhiding.com.
Whether it is Microsoft, Google, Amazon, or domestic BATH, a large number of engineers have invested in the field of artificial intelligence to develop AI algorithms for various scenarios. The underlying technologies of various scenarios are essentially consistent, which forms the advantages of IT companies in underlying technologies and AI algorithms.
However, whether the AI algorithm is good or not can often not directly determine the actual usage effect. From the AI algorithm to the actual production link, there is often a sword of Damocles called "Scene".
"The AI algorithm developed by IT companies is very knowledge-how when applied to factory workshops. How much role a good tool can play is often determined by the experience of experts."
Taking AI quality inspection as an example, it is not only the quality of AI algorithms that determine the accuracy of AI quality inspection recognition, but also many factors such as camera adjustment, image data feedback, factory on-site management mode, etc. "If your camera is The quality of the photos taken by the machine is unclear. No matter how powerful your AI algorithm is, it cannot reach a very high level through subsequent learning. These all determine that we actually understand the manufacturing scenarios better than IT technology companies. "
IT companies are better at the development of underlying algorithms. Therefore, more and more Internet industrial cloud platforms have been emerging in those years; teams from manufacturing backgrounds have a deeper understanding of the scenarios, so the real benchmark projects of intelligent manufacturing come from the intelligent manufacturing teams incubated within the traditional manufacturing industry.
Auto Digital is such an intelligent manufacturing team incubated by AUO Optoelectronics. After handling more and more intelligent manufacturing transformation projects, this team has gradually built its own industrial Internet cloud platform.
Industrial AI is in turmoil
On November 11, 2020, at the AUO Technology Trend Forum, AUO Digital released the industrial Internet cloud platform for the optoelectronics industry - Dazhi Oasis. This is the third year of AUO Digital's establishment and the last year of AUO's second three-year plan.
is different from the Internet cloud platform that builds the underlying technology created by IT companies. Zhao Lina told Zhiding.com, "We released an industrial Internet cloud platform for industry scenario applications at that time. The PaaS layer and IaaS layer will also use the AI algorithms and basic cloud services of Internet companies. What we mainly do is the functions and services of the PaaS layer and SaaS layer." After the accumulation of the past three years, AUD at that time had already had more than 500 customers across 23 industries. AUD will accumulate the algorithms, services and industry knowledge accumulated in these industry customer cases, as well as the cases and knowledge built by itself in the industrial ecosystem across 20+ industries and 300+ enterprises.
This makes Dazhi Oasis a universal industrial Internet cloud platform.
However, Zhao Lina also admitted, "Although some products on Dazhi Oasis are general products, most of them will still distinguish industries. For example, products like MES will still have different MES versions for different industries."
In the years of active transformation of intelligent manufacturing, IT companies are not willing to provide underlying cloud services. They once launched industry-oriented Internet cloud platforms or regional Internet cloud platforms in depth, and even hired a large number of industry experts to help themselves promote their business. However, when they were actually implemented, the expected results were not achieved. Where is the problem of
?
Zhao Lina found in promoting the intelligent manufacturing project that the biggest problem in the implementation of the intelligent manufacturing project is that most Party A customers cannot express their real needs. The work of finding out the needs of enterprises falls on smart manufacturing solution manufacturers. To understand the needs of enterprises' smart manufacturing transformation, business personnel themselves need to be industry experts.
"People who understand AI do not understand scenes, and those who understand scenes do not understand AI", this is still a true portrayal of the current rapidly developing industry. For IT companies, it is almost impossible to get technical experts to find demand in business scenarios.
A positive trend is that with the rapid development of artificial intelligence technology, the use of related tools has become easier and easier. As digital transformation becomes the main color of all industries, more and more business personnel in traditional enterprises have begun to step into the door of self-transformation, sounding the clarion call for the charge to artificial intelligence technology.
Zhao Lina said that in the process of technological development, there is usually a golden intersection between technological development and business applications. To promote the popularization of technology too early, it is often difficult to promote the business level. Only when this golden intersection arrives, as the cost of technology learning decreases, will a real technological reform storm be formed in the industry.
In the past few years, AUO alone has trained more than 800 AI experts and developed more than 2,000 AI models. Nearly 100 actual AI application cases have been found in 10 industries such as photoelectric panels, semiconductor IC design, petrochemical rubber, food and medicine, etc.
In the research and development of food and pharmaceutical formulas, AUD has been using AI algorithms to help customers analyze the impact of thousands of influencing factors on drug efficacy; in water quality detection, AUD has been using AI algorithms to judge sewage pollution...
such changes made Zhao Lina realize that the golden intersection is coming soon. Her judgment is that this year will be the first year for the real explosion of artificial intelligence technology in the manufacturing field.
However, she also told Zhiding.com, "Now, AUO Optoelectronics' third three-year plan around "5G+Industrial Internet" is also being promoted in an orderly manner. ”
