Manufacturing automation and AI might be here, but Industry 4.0 is still far off
The current social and economic issues surrounding manufacturing span everything from the ramifications for the American workforce to the international race towards technological dominance. Manufacturing has been a hot topic in politics for awhile now. For those of us working in the manufacturing space, however, these issues are nothing new. I recently sat down with Mark Jagiela, the CEO of Teradyne, a company that automates repetitive manual tasks and electronic tests for the semiconductor and electronics markets. Teradyne is the parent company of Universal Robots, a collaborative robotic system targeted at human-scale tasks in electronics assembly. During our conversation, we pinpointed one idea that no one has been talking about, but should be: the lagging adoption of automation technology across the manufacturing space – in the United States and elsewhere – and what that means for the future of manufacturing.
An understandable misunderstanding
From reading the headlines, it would be reasonable to have the misconception that automation was already in a rapid adoption phase – that is frankly not the case. While ubiquitous automation tools exist, like Surface Mount Assembly lines or Kuka robots assembling the body of cars, many would be surprised to learn just how many manual processes exist. For those of us in the space, the concept of Industry 4.0, where all of the machines will be connected together, is mythological: in many places, there aren’t even machines to connect.
Your cell phone or laptop? Yeah, those were built by hundreds, if not thousands of pairs of human hands. From assembly, to functional testing, to visual inspection, widespread use of humans and manual processes remain the status quo.
Are there robots that can do the jobs? According to Jagiela, yes – but it’s a bit more complicated than that. For years, the semiconductor industry has been using Teradyne automation for function testing. While other test station equipment exists on cell phone assembly lines, most of the loading and unloading is still very manual. Jagiela shared, “When we acquired Universal Robots we thought testing and robots would work together [just as it had for our semiconductor business]. It has been difficult to incorporate them onto the line.” From Jagiela’s perspective, the technology works – but the manufacturing ecosystem for consumer products is a complex multi-party system of misaligned incentives, which has been slowing adoption across the wider manufacturing space.
The reality of implementing automation technology
Companies like Teradyne work with the consumer electronics brands (whose logos are on the products) to understand the testing needs and work with a Systems Integrator to propose how it could work on the line. Ultimately though, they have to convince the manufacturing partner (the factory) to buy it.
There are two points of resistance from factory teams. First, the incentive structures in many of these organizations do not favor variation from the status quo – and in my experience, are sometimes incentivized only on cost reduction, not quality improvement. Jagiela shared, “Individual executives at these factories are reluctant to rock the boat. It tends to only be the enlightened factory owner that sees the total equation of quality and business impact of automation.” Second, there isn’t a universal standard for the calculation of the return on investment of automation, so the industry as a whole struggles with it.
I’ve seen this first hand while touring a factory building charging cables for a hot new wearable device. The automation team leader first showed me a bit of floorspace where seven human operators were assembling the cables. He then walked me into the next room where the same cables were being made with a fully automated process, and there were only two people working on the machine – one to feed parts in and one to pack the cables at the end of the line into a shipping carton. He explained, “When a customer sees seven people, they expect to pay for seven people. When a customer sees a robot, they expect to pay for no people. How can we recoup our cost?” As Jagiela puts it, “If the manufacturing partner doesn’t benefit monetarily, it is hard to justify. It is difficult to figure out who is motivated to make an automation investment happen [in these multi-party scenarios].”
The brands love the concept of automation because it aligns well with their incentives, believing it will increase product quality by reducing human error and increase the reliability of their supply chain because they will rely less on trained human workers who turnover. But there are many practical challenges, even with today’s advanced collaborative robots. While many assembly steps seem mind-numbingly repetitive when you are the person doing them, for a robot, many of them are still too difficult to automate due to the range of acceptable variations in the parts or the process steps. This variation was at the core of the Tesla Model 3 automation issues, and can be likened to all of my experiences assembling IKEA furniture: it’s not possible to do it without forcing at least some of the parts together, and that kind of insight and resulting part manipulation is a uniquely human ability. Another practicality that is slowing adoption is the complexity and risk in deploying this equipment, especially on lines that are already producing.
There are plenty of manufacturing tasks that are repetitive enough and where the incentives align. Jagiela shared that the average payback period for these kinds of applications for implementation in the U.S. is usually less than a year, and that he has multiple customers who have decided to spend more on automation than human labor would cost, because of the benefits of improved assembly quality.
Innovation for tomorrow
Automation has some hurdles to overcome to be able to win wide adoption. From a practical perspective, robotics need to continue to get better at on-the-fly adaptability, so that they can accomplish a greater range of tasks. From a business perspective, the best practice is to tie automation to core manufacturing metrics like yield, throughput, and efficiency which appeal to both brands and factories. For low-cost labor markets, it might be necessary to leverage the automation to add even more value that humans cannot provide such as data records, traceability, and adaptability. This is where technologies like Instrumental can help. Instrumental aggregates visual data and processes it with intelligent algorithms that not only detect unanticipated defects and speed up FA, but also provide a valuable, traceable data record that human inspection lacks. Instrumental also provides concrete ROI in the form of increased yield and efficiency, resulting in a cost savings of dollars per unit rather than pennies per unit. I’ve long believed that the future of manufacturing requires not just automation, but autonomy. Jagiela provided his perspective, “As robots deal with variation, autonomy will be key. They’ll end up with vision systems, neural networks, and machine learning that will allow them to do more bespoke tasks. Imagine that a defective product comes down the line, this robot will be able to adapt and make a custom product for the market. We will leave behind the draconian mass production of products.” Such a future would also have feedback loops and traceability not possible on today’s manual lines, which can recapture tremendous value out of the dollars wasted every day from inefficient assembly processes. To get there, both the buyers and the sellers of manufacturing automation need to align on this vision of the future and how to measure it.