It is no secret that one the biggest trends hitting the technology space is machine learning, promising new insights and bottom line benefits to organizations worldwide. The premise of machine learning is to allow machines to use large amounts of data to create algorithms in order to discover patterns and accurately predict future outcomes. This type of system creates a streamlined process that sidesteps the need to be programmed or reprogrammed specifically for every single possible action. 

The manufacturing industry stands to gain the most from taking advantage of these new capabilities, ushering in a new era of smart manufacturing. Semiconductor manufacturing, with its hundreds of precise steps and sensitive processes, is an ideal candidate to leverage machine learning. By using “connected” machines to collect real-time data in factories, manufacturers can dramatically improve their efficiency and productivity by making real-time decisions and predictions. 

A select group of industry and government leaders convened at the SEMICON West 2017 trade show on July 11 to discuss smart manufacturing technologies being pioneered by the semiconductor ecosystem. The participants are listed below, and the conversation that ensued was fascinating.

• Roberta Gamble, Vice President & Partner, Frost & Sullivan (moderator)
• Wayne Allan, Senior Vice President, Global Manufacturing, Micron Technology
• Tim Long, Enterprise Data Science, Micron Technology
• Tim Archer, Chief Operations Officer, LAM Research
• Maciej Kranz, Vice President, Strategic Innovations, Cisco Systems
• Gian Yi-Hsen, Regional President, North America Economic Development Board Singapore

Panel Highlights

Smart manufacturing is truly living up to its “smart” moniker largely because of one simple factor:  results. Wayne Allan explained how Micron's manufacturing data science teams, formalized in 2015 to harness data and analytics to serve top-line growth, have reported more than 2,800 data-science wins to date. The innovative big data infrastructure provided by IT's Enterprise Analytics and Data team has enabled a cross-functional partnership resulting in a 10 percent increase in product output — not a modest number when you consider Micron's 2016 revenues of $12.4 billion.  

Other data-science wins highlighted during the roundtable included Micron’s 35 percent reduction in quality-related excursions and the ability to yield targets 25 percent faster. Micron is not alone. According to Maciej Kranz, when Cisco installed 1,500 sensors in its Malaysian facility, the resulting data and analysis helped reduce the plant’s energy consumption by 30 percent. “It’s all about the data,” said Cisco’s Kranz. 

Simply said, smart manufacturing and machine learning strategies enable engineers to catch mistakes early when they’re cheaper to fix. It helps employees schedule and manage raw material stocks more efficiently, and it provides greater transparency to end-customers about product ship dates. As time goes on, more insights, efficiencies and savings are expected. “We’ve just started realizing the benefits,” said Micron’s Allan.

The Power of Partnering

One essential ingredient of smart manufacturing is partnering — between manufacturers and suppliers, between groups within companies and between companies and standards bodies. “No one organization has all the data it requires,” said Tim Archer, Chief Operating Officer of Lam Research. “Partnerships are needed” to help generate the data and then make use of it.

Illustrating this point, Lam now has nearly 1,000 sensors on a single manufacturing tool, and state-of-the-art fabrication facilities (or fabs) can have hundreds of such tools. This is one reason why, to date, Micron has collected more than 14 petabytes of manufacturing data from its 13 fabs.

Hear Micron SVP Wayne Allan discuss how big data analytics are used to improve yields and create a more efficient factory network. #smartMFG pic.twitter.com/XN4MZ8PNzD

— MicronTech (@MicronTech) August 8, 2017

Data analytics can now be used to get equipment ramped quickly and efficiently. In some cases, Lam has reduced the time needed to get tools fully functional from 21 days to less than a week, substantially reducing the time a facility needs to be ready for production. 

The sharing of data and insights between vendors, manufacturers and others is key but requires trust and IP protection. So far these collaborations seem to be working, but security and trust will continue to be key issues. “If data sharing breaks down, it could be a gating factor,” said Lam’s Archer.

Workforce Opportunity

Jobs inevitably change or are lost with the advent of new technologies. “This is always an issue during transformations,” said Yi-Hsen of Singapore’s Economic Development Board. “A key to managing such changes is ongoing worker education and training.” 

Singapore has worked for decades to keep its labor force at the cutting edge. It does this in part by working with manufacturers to develop university curricula and ongoing training. Recently this has meant adding analytics, data science and other technology-driven disciplines to training curricula. The ultimate, ongoing goal for Singapore is to compete by being among the most technologically advanced manufacturing locations.

Similarly, Cisco works with universities and other institutions to create curricula and internships, helping prepare students for new and emerging manufacturing jobs. “It’s a win-win because we get better candidates and schools graduate students with more relevant skills,” said Kranz of Cisco. “This is the right thing to do and it’s also in our best interest.”  

Even without explicit retraining, new opportunities emerge. By automating repetitive or mundane tasks like tool maintenance in a fab, time is freed to focus on more challenging and interesting problems — problems that an algorithm can’t yet crack. “We find that some of our best data scientists are former engineers,” said Long of Micron.

Manufacturing engineers can also play a key role in ramping new technologies. “Some of our engineers have decades of knowledge, and we include them in the process of designing new technologies,” said Cisco’s Kranz. “They are part of the solution.” 

What Lies Ahead

In increasingly competitive markets, improving manufacturing with data analytics is essential. Nowhere is this truer than in semiconductor manufacturing, where pressure mounts every year to reduce costs, improve efficiency and focus on the quality of mission-critical products. 

While the focus of this discussion was on smart manufacturing, the panel also touched on the dramatic impact the IoT, machine learning and data analytics have on other areas of the business. Cisco’s Kranz summed it up well when he talked about the impact of the IoT taking off about 10 years ago when “the line of business began becoming the primary consumer of the benefits of the connected world and IoT,” said Kranz. “Today, every company is becoming a technology company.”

You can watch the entire panel discussion below.