When governments opine about the loss of manufacturing jobs and talk about bringing those jobs back, they are really romanticising “the good old days”. Manufacturing today isn’t about hundreds or thousands of workers standing at production lines. It’s about skilled technicians, automation and being data-driven. I spoke with Seagate’s Jeff Nygaard about the new wave of manufacturing at the company’s factories.
Over the last two decades, Seagate has transitioned from highly manual production and now boasts a highly automated production line. But, unlike many of the doomsayers saying jobs will disappear as a result of automation, Nygaard said the opposite is true with Seagate hitting more people into their North American and European platter manufacturing centres.
What’s changing are the tasks those people are doing. Instead of assembling the complex hardware, they are now managing the machines that make the hard drives we rely on.
The learning factory
A major component of this process is the use of machine learning. While the use of sensors to measure machine performance in factories isn’t new, Nygaard says what’s changed is an increased dependence on unstructured data.
Older sensors provided lots of numerical data that was relatively easy tot process. But Seagate now uses cameras too photograph each component of a drive through the various steps in the manufacturing process – there are as many as 3000 steps in the creation of some hard drive models.
Nygaard said the company ships about 30 million drives per quarter. Each of those includes five or six platters, potentially two heads for reading and writing data as well as hundreds of other moving parts. Form start to end, the process takes about six months with the production of wafers including as many as 2000 process steps alone.
For example, Seagate’s Normandale factory takes 17 million microscope pictures every day, generating 10TB of data that’s analysed to detect potential production defects before wafers are assembled into drives. There are 100,000 sliders in every 200mm wafer that need to be checked.
There’s simply no way human operators can keep up with the volume of data they’d need to look through to detect a potential problem.
Moving towards automation
In the past, only a sample of the images were analysed using rudimentary algorithms and human analysts. But the company has invested in new machine learning and big data tools that allow it check every single device through the manufacturing process.
We use SSDs or flash storage at end points and hard drives for mass storage where performance isn’t quite as critical. However, a number of new technologies are emerging that tell us the hard drive will still be a big deal for many years to come. Here are some of the biggest innovations we’ll see in the coming months and years.”]
That greatly reduces the potential for a flawed component making it to a later stage in the process which can lead to greater disruption and cost.
The journey towards this automated process required several shifts for Seagate. Nygaard said there was a need to “sensorise” the plants in order to monitor environmental conditions such as temperature, humidity and air pressure as well as the use of cameras to capture images and video of machines as they worked as well as products as they are assembled.
People still matter – a lot
There’s an obvious impact on the people who work at Seagate’s manufacturing plants. But Nygaard said the company needs more people, not fewer. It is investing in training and education, equipping people to become “citizen data scientists” who can help create new algorithms and tune the learning system to become more accurate at detecting anomalies.
He also notes that there’s still a lot of work for people to do in Seagate’s factories. But the work content has changed. Rather than making parts using “the dexterity of their fingers” workers are now managing the tools that make and assemble each of the parts.
The changes Seagate has made have been incremental – “evolutionary rather that revolutionary” said Nygaard. That’s allowed the company to bring people along for the journey. There’s also a strong focus on creating and fostering relationships with local vocational schools.
Nygaard added that this isn’t a task with an end in sight. There is a long-term commitment to continually improve the processes and systems.
Sitting behind all this is Project Athena. This is a three-piece initiative that brings together a data lake, inference engine and a deep learning neural network that supports the continuous improvement of the manufacturing process. It uses the data to help make real-time decisions as hard drives are produced.
Seagate said that while Project Athena may be excellent at identifying defects, it doesn’t completely replace factory subject-matter experts. And it’s unlikely that it ever will. It said Project Athena opens up new opportunities for Seagate’s human wafer experts to innovate and remedy larger problems.