Lights-Out Automation: Creating Resilient Factories

There’s a very well-known joke about the perfect factory: “It runs with only one human and one dog. The human is there to feed the dog, and the dog is there to prevent the human from touching anything.”

Lights-out manufacturing refers to factories that run fully automated and thus require no humans on-site. As no humans are present, the lights can be turned off (or there can even be no lights) and hence this type of manufacturing is also called also called “dark factory.” The concept of lights-out factories started in 1955 and was given its first real-life test by General Motors in 1982.

Humans in work

For most of the history of mankind, work consisted essentially of human physical labor (even if sometimes supplemented with animal physical labor).

Humans are formidable workers. They have great dexterity, versatility, adaptability and above all, great intelligence. However, humans also have limitations: they can only perform tasks at a certain speed and with certain precision; they sometimes make mistakes; they get bored with repetitive tasks and they also have moods. According to the Maslow Hierarchy of Needs, humans have increasing levels of needs, starting with the base of the hierarch, physiological needs (water, food, rest, shelter, clothing); then safety needs (health, personal security, employment, property); then love and belonging (friendship, intimacy, family); then esteem (respect, self-esteem, status, recognition); and at the top of the hierarchy, self-actualization (desire to realize one’s potential).

Machines on the other hand can perform tasks at a very high-speed with very high precision. Machines don’t have moods (yet) and can work 7 x 24 hours without degradation in their level of work. Machines do not have a Hierarchy of Needs, although they need power and other supplies (e.g. water, gas, lubrication) and they also need maintenance.

When the industrial revolution started at the end of the 18^th century in England, the first machines were quite rudimentary and were applied to very repetitive tasks in the textile industry. The productivity gains were extraordinary: much more goods could be produced with the same (or less) inputs. The productivity gains drove costs and prices down, which in turn significantly increased demand. What about the effect on the number of workers in the textile industry? In 1760 in England, before the invention of the cotton-spinning machine by Richard Arkwright there were 7,900 people engaged in the production of cotton textiles. Due to the increase in demand, by 1787 the number of persons engaged in the production of cotton textiles had risen to 320,000– an increase of 4,400% in the span of 27 years [1]. Most of the labor in this case came from people that lived in the countryside and worked in agriculture.

This effect of technology progress driving work displacement is not new. However, the net effect for both the society and employment as a whole was always positive, even if some sectors experienced some short-term pain. With every new technology there is the fear of job loss. Still, in 2020, after all the technology and automation advances, there have never been more people employed in the history of mankind.

The Effect of Industrial Automation

There’s a saying that “necessity is the mother of invention.” Historically, the industries that led the way in terms of automation were industries that were initially labor intensive and where there were a lot of repetitive tasks. It started in the textile industry, and progressed quickly to automotive production, and more recently in semiconductor and electronics.

The first MES as well as automation and integration standards were created specifically for the semiconductor industry. Today, the semiconductor factory is far more advanced, more standardized and the most automated industry in the world.

Today, the drive for automation is omnipresent. The International Federation of Robotics forecasted that more than 1.7 million new robots would be installed in factories worldwide by 2020 [2]. Interestingly enough, the driver for automation today is not only the reduction of the labor costs but also the production efficiency that comes with maximizing equipment utilization, especially in capital-intensive industries.

There have been many technological and standardization advances that make automation possible. Still, costs aside, there are two types of complex tasks that are very hard to automate: 1) material handling and logistics tasks and 2) tasks which are complex, which require special dexterity and include some degree of randomization (see Elon Musk’s comments “humans are underrated” [3]).

Continuous process industries, such as oil refining and power stations, are relatively easily to automate because they don’t experience the two types of complex tasks stated above. Conversely, automation is particularly challenging in the discrete manufacturing industry.

Regarding material handling and logistics tasks, while it’s very hard to automate the inbound and outbound movements of material, great strides have been made in automating the internal logistics, both in the warehouses and on the shop floor with Automated Storage Systems, Automated Guided Vehicles (AGVs) and Autonomous Intelligent Vehicles (AIVs). It’s important to emphasize that the more standardized the material is, the easier it is to automate. In this regard, good examples of standardization are semiconductors and Surface Mount Technology (SMT) manufacturing.

Regarding the tasks which are complex, require special dexterity and include some degree of randomization, there have been great advances in using robotics, computer vision, computing, Internet of Things and machine learning for task automation.  

A key enabler of highly automated manufacturing for the high tech industries is a factory-level automation controller that supervises, coordinates and orchestrates the several factory systems and sub-systems such as MES, ERP and PLM applications as well as Process, Metrology and Logistics Equipment. This factory automation controller is responsible for:

1. Reacting to different factory events that the systems generate (e.g. job completed; new order arrived; equipment down)
2. Directing the execution of the different systems (e.g. transport lot L1 from location A to location B; return raw material M1 to warehouse)
3. Keeping track of the state of every activity in the factory
4. Handling errors and exceptions  

The handling of errors and exceptions presents the greatest challenge, as some of the errors can be unforeseen and the cost of automating the error handling can be prohibitively expensive.

This state-of-the-art in automation can be found in semiconductor front-end fabrication, but these types of factories are far from being a lights-out fab. People can still be found, but they are typically 1) located in a control room, checking the state of the factory by looking at the factory-level automation controller dashboard and handling exceptions manually; and 2) performing preventive or corrective maintenance on equipment.

There are many companies that while not as automated as a semiconductor front-end are already using automatic transport systems with operators handling only the handoffs between the transport system and the equipment and vice-versa. What is interesting in this case is that the same automation architecture with the factory-level automation controller can be used as well. Similarly, the same control room concept can be maintained to monitor the performance of the factory as well to look for errors and exceptions that need human intervention.

Benefits of a Highly Automated, Resilient Factory

A highly automated factory has many benefits:

• Increased productivity and quality
• Improved consistency and predictability
• Decreased number of work-related accidents
• People used for higher, value-add activities
• Less dependency on local labor, so factories can be located closer to the end user and/or R&D, reducing both the transport times and overall production costs

A highly automated factory is also inherently more resilient. In times of crisis, such as with the Coronavirus COVID-19 epidemic that the world is experiencing today, a highly automated factory is far more likely to continue manufacturing operations than a factory with low automation. It is also able to ramp production to respond to demand.  In situations such as a health crisis, it’s particularly important to ensure that essential goods, such as healthcare-related products (ventilators, masks and hand sanitizers) can be manufactured at high volumes without the need to put the workers at risk of infection, especially considering that some workers may be sick or quarantined.

While a complete lights-out manufacturing factory may still be several years away, the companies that are best positioned to survive and thrive in the future are the ones that embark on the journey toward full automation, taking decisive steps, one step at a time.

[1] Economics in One Lesson, by Henry Hazlitt (https://mises.org/library/economics-one-lesson)

[2] IFR forecast: 1.7 million new robots to transform the world´s factories by 2020 (https://ifr.org/ifr-press-releases/news/ifr-forecast-1.7-million-new-robots-to-transform-the-worlds-factories-by-20)

[3] Elon Musk says “humans are underrated” after his robots slow Model 3 production (https://www.fastcompany.com/40559386/elon-musk-says-humans-are-underrated-after-his-robots-slow-model-3-production)