From advanced robotics in R&D labs to computer vision in warehouses, technology is making an impact on every step of the manufacturing process.
“Lights-out manufacturing” refers to factories that operate autonomously and require no human presence. Because they don’t need human supervision, they don’t require lighting, and can consist of several machines functioning in the dark.
While this may sound like science fiction, these kinds of factories have been a reality for more than 15 years.
The Japanese robotics maker FANUC has been operating a “lights-out” factory since 2001, where robots build other robots completely unsupervised for nearly a month at a time.
“Not only is it lights-out,” said FANUC VP Gary Zywiol, “we turn off the air conditioning and heat too.”
To imagine a world where robots do all the physical work, one simply needs to look at the most ambitious and technology-laden factories of today.
In June 2018, the Chinese e-commerce giant JD.com unveiled a fully automated storage and shipping facility in Shanghai.
The factory is outfitted with twenty industrial robots that can pick, pack, and transfer packages with no human presence or oversight.
Without robots, it would take as many as 500 workers to fully staff this 40K square foot warehouse — instead, the factory requires only five technicians to service the machines and keep them working.
As industrial technology grows increasingly pervasive, this wave of automation and digitization is being labelled “Industry 4.0,” as in the fourth industrial revolution.
So, what does the future of factories hold?
To answer this, we took a deep dive into 8 different steps of the manufacturing process, to see how they are starting to change:
- Product R&D: A look at how platforms are democratizing R&D talent, the ways AI is helping materials science, and how the drafting board of tomorrow could be an AR or VR headset.
- Resource Planning & Sourcing: On-demand decentralized manufacturing and blockchain projects are working on the complexities of integrating suppliers.
- Operations Technology Monitoring & Machine Data: A look at the IT stack and platforms powering future factories. First, factories will get basic digitization, and further along we’ll see greater predictive power.
- Labor Augmentation & Management: AR, wearables, and exoskeletons are augmenting human capabilities on the factory floor.
- Machining, Production & Assembly: Modular equipment and custom machines like 3D printers are enabling manufacturers to handle greater demand for variety.
- Quality Assurance (QA): A look at how computer vision will find imperfections, and how software and blockchain tech will more quickly be able to identify problems (and implement recalls).
- Warehousing: New warehouse demand could bring “lights-out” warehouses even faster than an unmanned factory, with the help of robotics and vision tracking.
- Transport & Supply Chain Management: Telematics, IoT, and autonomous vehicles will bring greater efficiency and granularity for manufacturers delivering their products.
Despite representing 11.6% of US GDP, manufacturing remains an area of relatively low digitization — meaning there’s plenty of headroom for automation and software-led improvements. In fact, in 2017, 76% of manufacturers reported having a smart factory initiative in the works.
Manufacturing is deeply changing with new technology, and nearly every manufacturing vertical — from cars, to electronics, to pharmaceuticals — is implicated. The timelines and technologies will vary by sector, but most steps in nearly every vertical will see improvement.
Read on for a closer look at how technology is transforming each step of the manufacturing process.
1. Product R&D
From drug production to industrial design, the planning stage is crucial for mass-production. Across industries, designers, chemists, and engineers are constantly hypothesis testing.
Will this design look right? Does this compound fit our needs? Testing and iterating is the essence of research and development. And the nature of mass-production makes last-minute redesigns costly.
Major corporations across drugs, technology, aerospace, and more pour billions of dollars each year into R&D.
In the highly-scientific world of R&D, high-caliber talent is distributed across the globe. Now, software is helping companies tap into that pool.
When it comes to networking untapped talent in data science and finance, platforms like Kaggle, Quantopian, and Numerai are democratizing “quant” work and compensating their collaborators. The concept has also already taken off with pharmaceutical R&D, though it’s growing elsewhere as well. On-demand science platforms like Science Exchange are currently working across R&D verticals, and allow corporations to quickly solve for a lack of on-site talent by outsourcing R&D.
While R&D scientists may seem non-essential to the manufacturing process, they are increasingly critical for delivering the latest and greatest technology, especially in high-tech manufacturing.
Companies are exploring robotics, 3D printing, and artificial intelligence as avenues to improve the R&D process and reduce uncertainty when going into production. But the process of hypothesis testing has room for improvement, and tightening iteration time will translate to faster and better discoveries.
Robotics & 3D printing speed up product development across verticals
Accelerating product development is the #1 priority for firms using 3D printing, according to a recent industry survey.
Moreover, most 3D printing use is directed at prototyping new technology.
3D printing is already a staple in any design studio. Before ordering thousands of physical parts, designers can us 3D printing to see what a future product looks like.
Similarly, robotics is automating the physical process of trial-and-error across a wide array of verticals.
In R&D for synthetic biology, for example, robotics is making a big impact for companies like Zymergen and Ginkgo Bioworks, which manufacture custom chemicals from yeast microbes. Finding the perfect microbe requires testing up to 4,000 different variants concurrently, which translates to lot of wet lab work.
Using automatic pipette systems and robotics arms, liquid handling robots permit high-throughput experimentation to arrive at a winning combination faster and with less human error.
Below is the robot gene tester Counsyl (left), used for transferring samples, and Zymergen’s pipetting robot (right) for automating microbe culture testing.
“Materials engineering is the ability to detect a very small particle — something like a 10-nanometer particle on a 300-millimeter wafer. That is really equivalent to finding an ant in the city of Seattle.” — Om Nalamasu, CTO at Applied Materials
Companies are now emerging to make these kinds of automatic pipetting technologies and others more accessible. Y Combinator alum Opentrons has raised nearly $30M for its platform designed to allow scientists to build the logic to automate repetitive experiments without using code — it claims to have 90% of the top 50 research universities now using its software and robots.
Its $4,000 lab robot, the OT-2, comes with a library of pre-programmed experimental routines that researchers can use to build their own protocols.
Looking beyond biotech, material science has played a pivotal role in computing and electronics.
Notably, chip manufacturers like Intel and Samsung are among the largest R&D spenders in the world. As semiconductors get ever-smaller, working at nanoscale requires precision beyond human ability, making robotics the preferred option.
Tomorrow’s scientific tools will be increasingly more automated and precise to handle micro-scale precision.
AI is hastening materials science discoveries
Currently, the hottest area for deals to AI startups is healthcare, as companies employ AI for drug discovery pipelines. Pharma companies are pouring cash into startups tracing drug R&D such as Recursion Pharmaceuticals and twoXAR, and it’s only a matter of time until this takes off elsewhere.
One company working in chemistry and materials science is Citrine Informatics (below, left). Citrine runs AI on its massive materials database, and claims it helps organizations hit R&D and manufacturing milestones 50% of the time. In 2018, Citrine raised $8M from Tencent to aid its international expansion — the next year, it announced a partnership with LANXESS to work on the use of AI for plastics production. Similarly, Deepchem (right) develops a Python library for applying deep learning to chemistry.
In short, manufacturers across sectors — industrial biotech, drugs, cars, electronics, or other material goods — are relying on robotic automation and 3D printing to remain competitive and tighten the feedback loop in bringing a product to launch.
Already, startups developing or commercializing complex materials are taking off in the 3D printing world. Companies like MarkForged employ carbon fiber composites, where others like BMF Material Technology are developing composites with rare nanostructures and exotic physical properties. As of March 2019, MarkForged has provided 3D printing services to Google, Amazon, and General Motors, and the company shipped 2,500 printers in 2018.
Certainly, manufacturers of the future will be relying on intelligent software to make their R&D discoveries.
Augmented and virtual reality ‘abstract away’ the modeling process
Currently, manufacturers of all types rely on prototyping with computer aided design (CAD) software. In future manufacturing processes, augmented and virtual reality could play a greater role in R&D, and could effectively “abstract away” the desktop PC for industrial designers, possibly eliminating the need for 3D printed physical models.
Autodesk, the software developer of AutoCAD, is a bellwether for the future of prototyping and collaboration technology. The company has been no stranger to investing in cutting-edge technology such as 3D printing, including a partnership with health AI startup Atomwise. Recently, Autodesk’s exploration into making an AR/VR game engine foreshadows the larger role it envisions for immersive computing in the design process.
Autodesk’s game engine, called Stingray, has support for the HTC Vive and Oculus Rift headsets. Additionally, game and VR engine maker Unity has a partnership with Autodesk to increase interoperability.
Similarly, Apple has imagined AR/VR facilitating the design process in combination with 3D printing. Using the CB Insights database, we surfaced an Apple patent that envisions AR “overlaying computer-generated virtual information” onto real-world views of existing objects, effectively allowing industrial designers to make 3D-printed “edits” onto existing or unfinished objects.
The patent envisions using AR through “semi-transparent glasses,” but also mentions a “mobile device equipped with a camera,” hinting at potential 3D printing opportunities for using ARKit on an iPhone.
A researcher at Cornell has recently demonstrated the ability to sketch with AR/VR while 3D printing. Eventually, the human-computer interface could be so seamless that 3D models can be sculpted in real time.
Tomorrow’s R&D team will be exploring AR and VR, and testing how it works in combination with 3D printing, as well as the traditional prototyping stack.
2. Resource Planning & Sourcing
Once a product design is finalized, the next step is planning how it will be made at production scale. Typically, this requires gathering a web of parts suppliers, basic materials makers, and contract manufacturers to fulfill a large-scale build of the product. But finding suppliers and gaining trust is a difficult and time-consuming process.
The vacuum maker Dyson, for example, took up to two years to find suppliers for its new push into the auto industry: “Whether you’re a Dyson or a Toyota it takes 18 months to tool for headlights,” a worker on their project reported.
Assembly lines today are so lean they’re integrating a nearly real-time inflow of parts and assembling them as fast as they arrive. Honda’s UK-based assembly factory, for example, only keeps one hour’s worth of parts ready to go. After Brexit, the company reported longer holdups for incoming parts at the border, and said that each 15 minute delay translates to £850,000 per year.
We looked at how technology is improving this complicated sourcing process.
Decentralized parts manufacturing
Decentralized manufacturing may be one impending change that helps manufacturers handle demand for parts orders.
Distributed or decentralized manufacturing employs a network of geographically dispersed facilities that are coordinated with IT. Parts orders, especially for making medium- or small-run items like 3D printed parts, can be fulfilled at scale using distributed manufacturing platforms.
Companies like Xometry and Maketime offer on-demand additive manufacturing and CNC-milling (a subtractive method that carves an object out of a block), fulfilling parts orders across its networks of workshops.
Xometry’s site allows users to simply upload a 3D file and get quotes on milling, 3D printing, or even injection molding for parts. The company is also working on CAD integration to simplify the process of ordering. To fulfill all those on-demand orders, the company works with more than 3,000 different materials suppliers. Xometry raised $50M from Dell, BMW, and GE (among others) in 2019.
The similar Amsterdam-based 3D Hubs promises the ability to take a quote on a part and put it into production “in less than 5 minutes.” In 2018, 3D Hubs transitioned from its original business model as a community 3D printer to a focus on high-end plastic, metal, and injected-molding manufacturing.
Xometry and 3D Hubs aren’t alone in offering printing services: UPS is also embracing the movement, offering services for 3D printed plastic parts like nozzles and brackets in 60 locations and using its logistics network to deliver orders globally.
As mass-customization takes off, so could the reliance on decentralized network of parts suppliers.
Blockchain for resource tracking
Enterprise resource planning (ERP) software tracks resource allocation from raw material procurement all the way through customer relationship management (CRM).
Yet a manufacturing business can have so many disparate ERP systems and siloed data that, ironically, the ERP “stack” (which is intended to simplify things) can itself become a tangled mess of cobbled-together software.
In fact, a 2017 PwC report found that many large industrial manufacturers have as many as 100 different ERP systems.
Blockchain and distributed ledger technologies (DLT) projects aim to unite data from a company’s various processes and stakeholders into a universal data structure. Many corporate giants are piloting blockchain projects, often specifically aiming to reduce the complexity and disparities of their siloed databases.
In 2017, British Airways tested blockchain technology to maintain a unified database of information on flights and stop conflicting flight information from appearing at gates, on airport monitors, at airline websites, and in customer apps.
When it comes to keeping track of the sourcing of parts and raw materials, blockchain can manage the disparate inflows to a factory. With blockchain, as products change hands across a supply chain from manufacture to sale, the transactions can be documented on a permanent decentralized record — reducing time delays, added costs, and human errors.
Treum, a project out of the Ethereum-based startup studio Consensys, works on a number of capital-intensive areas that serve manufacturers. And Provenance is building a traceability system for materials and products, enabling businesses to engage consumers at the point of sale with information gathered collaboratively from suppliers all along the supply chain.
Going forward, we can expect more blockchain projects to build supply chain management (SCM) software, handle machine-to-machine (M2M) communication and payments, and promote cybersecurity by keeping a company’s data footprint smaller.
3. Operations Technology: Monitoring & Machine Data
Industrial downtime is a $647B a year problem, according to the International Society of Automation.
Presumably, tomorrow’s manufacturing process will eventually look like one huge, self-sustaining cyber-physical organism that only intermittently requires human intervention. But across sectors, the manufacturing process has a long way to go before we get there.
According to lean manufacturing metrics (measured by overall equipment effectiveness, or OEE), world-class manufacturing sites are working at 85% of their theoretical capacity. Yet the average factory is only at about 60%, meaning there’s vast room for improvement in terms of how activities are streamlined.
Industry 4.0’s maturation over the next two decades will first require basic digitization.
Initially, we’ll see a wave of machines become more digital-friendly. Later, that digitization could translate into predictive maintenance and true predictive intelligence.
Large capital goods have evolved to a “power by the hour” business model that guarantees uptime. Power by the hour (or performance-based contracting) is now fairly common in the manufacturing world, especially in mission-critical areas like semiconductors, aerospace, and defense.
The idea dates back to the 1960s, when jet engine manufacturers like GE Aviation, Rolls Royce, and Pratt & Whitney began selling “thrust hours,” as opposed to one-off engine sales. This allows engine makers to escape the commodity trap and to focus on high-margin maintenance and digital platforms. Nowadays, GE is incentivized to track every detail of its engine, because it only gets paid if the engine is working properly.
Despite a guarantee of uptime, a machine’s owner is responsible for optimizing usage (just like airlines that buy jet engines still need to put them to good use). In short, factory owners still “own” the output risk between the chain of machines.
Without digitizing every step, efficiency is being left on the table. Yet there are serious barriers for manufacturers to take on the new burden of analytics.
Shop floors typically contain old machines that still have decades of production left in them. In addition to significant cost, sensors tracking temperature and vibration aren’t made with a typical machine in mind, lengthening the calibration period and efficacy.
When Harley-Davidson’s manufacturing plant went through an IIoT sensor retrofit, Mike Fisher, a general manager at the company, said sensors “make the equipment more complicated, and they are themselves complicated. But with the complexity comes opportunity.”
From initial digitization to predictive
To put it simply, operational technology (or OT) is similar to traditional IT, but tailored for the “uncarpeted areas.” Where the typical IT stack includes desktops, laptops, and connectivity for knowledge work and proprietary data, OT manages the direct control or monitoring of physical devices.
For manufacturers, the OT stack typically includes:
- Connected manufacturing equipment (often with retrofitted industrial IoT sensors)
- Supervisory control and data acquisition (SCADA) systems and human machine interfaces (HMI), which provide industrial monitoring for operations analysts
- Programmable logic controllers (PLCs), the ruggedized computers that grab data on factory machines
- 3D printers (additive manufacturing) and computer numerical control (CNC) machines for subtractive manufacturing (like whittling away a block)
In a way, IT and OT are two sides to the same tech stack token, and as manufacturing gets better digitized, the boundaries will continue to blur.
Today, the “brain” for most industrial machines is in the programmable logic controller (PLC), which are ruggedized computers. Industrial giants like Siemens, ABB, Schneider, and Rockwell Automation all offer high-priced PLCs, but these can be unnecessarily expensive for smaller manufacturing firms.
This has created an opportunity for startups like Oden Technologies to bring off-the-shelf computing hardware that can plug into most machines directly, or integrate existing PLCs. This, in turn, allows small- and medium-sized businesses to be leaner and analyze their efficiency in real time.
As digitization becomes ubiquitous, the next wave in tech efficiency improvements will be about predictive analytics. Today’s narrative around the Internet of Things has suggested that everything — every conveyor and robotic actuator — will have a sensor, but not all factory functions are of equal value.
Slapping cheap IoT sensors on everything isn’t a cure-all, and it’s entirely possible that more value gets created from a smaller number of more specialized, highly accurate IoT sensors. Augury, for example, uses AI-equipped sensors to listen to machines and predict failure.
Cost-conscious factory owners will recognize that highly accurate sensors will deliver greater ROI than needless IoT.
New architecture at the edge
Computing done at the “edge,” or closer to the sensor, is a new trend within IIoT architecture.
Drafting on innovations in AI, and smarter hardware, Peter Levine of a16z anticipates an end to cloud computing for AVs, drones, and advanced IoT objects.
Connected machines in future factories should be no different.
Edge computing offers significant benefits to manufacturers, including:
- Increased efficiency: Manufacturers can process their data close to the source where it is captured, allowing applications to operate faster
- Lower costs: Manufacturers can avoid expensive cloud storage and processing fees, while processing their data on inexpensive local devices
- Efficient bandwidth: As cloud usage increases, edge computing allows manufacturers to avoid competing for increasingly sparse and expensive bandwidth
Companies like Saguna Networks specialize in edge computing (close to the point of collection), whereas a company like Foghorn Systems does fog computing (think a lower-hanging cloud that’s done on-site like a LAN). Both methods allow mission-critical devices to operate safely without the latency of transmitting all data to a cloud.
In the near future, advances in AI and hardware will allow IoT as we know it to be nearly independent of centralized clouds.
This is important because in the short term, it means that rural factories don’t need to send 10,000 machine messages relaying “I’m OK,” which expends costly bandwidth and compute. Instead, they can just send anomalies to a centralized server and mostly handle the decision-making locally.
Additionally, cloud computing latency has drastic downsides in manufacturing. Mission critical-systems such as connected factories can’t afford the delay of sending packets to off-site cloud databases. Cutting power to a machine split-seconds too late is the difference between avoiding and incurring physical damage.
And in the longer term, edge computing lays down the rails for the autonomous factory. The AI software underpinning the edge will be the infrastructure that allows factory machines to make decisions independently.
In sum, devices that leverage greater computing at the edge of the network are poised to usher in a new, decentralized wave of factory devices.
Cybersecurity is a priority
One paradox of IIoT is that factories bear significant downside risk, yet are barely investing in protection. While research has shown that cyber espionage affects manufacturing more than any other industry, and 81% of organizations are concerned about the security risks of factory floor IoT, only 37% of corporations believe their internal cybersecurity awareness programs are effective. Further, 48% of manufacturers have suffered some kind of cybersecurity incident.
Cyber attacks can be devastating to heavy industry, where cyber-physical systems can be compromised. The WannaCry ransomware attack caused shutdowns at the Renault-Nissan auto plants in Europe. In 2019, another ransomware attack on the Norwegian aluminum producer Norsk Hydro resulted in estimated losses of $41M of lost production capacity.
Consequently, critical infrastructure is a growing segment within cybersecurity, and many startups like Bayshore Networks are offering IoT gateways (which bridge the disparate protocols for connected sensors) to allow manufacturers across many verticals to monitor their IIoT networks. Other gateway-based security companies like Xage are even employing blockchain’s tamperproof ledgers so industrial sensors can share data securely.
81% of organizations are concerned about the security risks of factory floor IoT, yet only 37% of corporations believe their internal cybersecurity awareness programs are effective.
Similarly, adding connected IoT objects and Industrial Control System (ICS) sensors has opened up new vulnerabilities at the endpoint.
Companies like Rubicon Labs and Mocana are developing secure communication products at the IP and the device level.
Mocana sells end-to-end cybersecurity suites specialized for IoT devices, with customers like Samsung, Verizon, Xerox and Panasonic. In 2019, the company raised $15M to expand into visibility and analytics tools.
Additionally, several of the most active enterprise cybersecurity investors are corporates with interests in OT computing. The venture arms of Dell (which makes industrial IoT gateways), as well as Google, GE, Samsung, and Intel are among the most active in this space.
Managing the ICS and IIoT systems securely will continue to be a critical area for investment, especially as hack after hack proves OT’s vulnerability.
4. Labor Augmentation & Management
In a 2017 write-up about furniture maker Steelcase’s production line, humans were described as being solely present to guide automation technology.
Steelcase’s “vision tables,” which are computerized workstations that dictate step-by-step instructions, eliminate human error in assembling furniture. Using sound cues and overhead scanners to track assembly, the system won’t let workers proceed if a step is done incorrectly. Scanners also allow off-site operations engineers to analyze progress in real time.
The New Yorker wrote about Steelcase’s labor management, “A decade ago, industrial robots assisted workers in their tasks. Now workers — those who remain — assist the robots in theirs.”
What manufacturing looks like has changed drastically in a short time. As a retired Siemens executive recently said, “People on the plant floor need to be much more skilled than they were in the past. There are no jobs for high school graduates at Siemens today.”
But better digitization and cyber-physical technologies are all augmenting the efficiency and manpower available to the workers. Here’s how emerging technology like augmented reality (AR), wearables, and exosuits are fitting in.
AR and mobile are digitizing the instruction manual
Augmented reality will be able to boost the skills of industrial worker.
In addition to being a hands-free “browser” that can communicate factory performance indicators and assign work, AR can analyze complicated machine environments and use computer vision to map out a machine’s parts, like a real-time visual manual. This makes highly skilled labor like field service a “downloadable” skill (in a manner not unlike The Matrix).
Daqri and Atheer are well-funded headset makers that focus on industrial settings. Upskill‘s Skylight platform (below) makes AR for the industrial workforce using Google Glass, Vuzix, ODG, and Realwear headsets. The company has raised nearly $50M from the corporate venture arms of Boeing and GE, among other investors.
Many AR makers envision the tech working like a handsfree “internet browser” that allows workers to see real-time stats of relevant information. Realwear‘s wearable display doesn’t aspire to true augmented reality like a Daqri headset, but even a small display in the corner of the eye is fairly robust.
Others like Scope AR do similar work in field service using mobile and iPad cameras, employing AR to highlight parts on industrial equipment and connecting to support experts in real time. This saves on the travel costs of flying out people to repair broken equipment.
Re’flekt, an enterprise AR developer from Munich, has built a platform for turning CAD data into augmented reality applications for maintenance and training. Jaguar Land Rover used REFLEKT ONE to build a training app that would allow employees to acquire “X-ray vision” into a car and let them identify the exact component or repair needed.
As the maxim goes, “what gets measured gets managed,” and in an area where robots are a constant competitive pressure, manufacturing organizations will invest in technologies that digitize human efforts down to each movement.
Exosuits & safety tech will become standard in dirty & dangerous jobs
Exoskeleton technology is finally becoming a reality on factory floors, which could drastically reduce the physical toll of repetitive work. Startups here are making wearable high-tech gear that bear the load alongside a worker’s limbs and back.
Ekso Bionics, seen below, is piloting its EksoVest suit at Ford Motor Company’s Michigan assembly plants, and workers using the suit have reported less neck stress in their daily demands. The EksoVest reduces wear from repetitive motion and, unlike some competing products, provides lift assistance without batteries or robotics. Ekso’s CTO has said the long-term strategy is to get workers accustomed to the technology before eventually moving into powered exoskeletons.
Sarcos is another well-known exosuit maker, which has raised from corporates including Schlumberger, Caterpillar, and Microsoft and GE’s venture arms. Sarcos is more strictly focused on remote controlled robotics and powered exoskeletons. Its robotic exoskeleton, which a worker can put on or take off in 30 seconds, can help a user lift and put down 200 lbs repeatedly for an up to eight-hour work session. In 2018, Delta became one of the initial members of Sarcos’s Exoskeleton Technical Advisory Group (X-TAG) alongside Bechtel and BMW.
In similar territory is Strong Arm Technologies, which makes posture-measuring and lift-assisting wearables. Strong Arm touts predictive power to intervene before risk of injury or incident, and is positioned as a labor-focused risk management platform.
Where humans are still needed for some dirty and dangerous tasks, wearables and exoskeletons will augment human’s ability to do work while also promoting safety.
5. Machining, Production, & Assembly
Automation is coming for dirty, dull, and dangerous jobs first.
Already, many human jobs within the mass-production assembly line have been crowded out by automation. Cyber-physical systems like industrial robotics and 3D printing are increasingly common in the modern factory. Robots have gotten cheaper, more accurate, safer, and more prevalent alongside humans.
Consumer tastes have also broadened, and manufacturers are trying to keep up with increasing demands for customization and variety.
Visions for Industry 4.0 involve a completely intelligent factory where networked machines and products communicate through IoT technology, and not only prototype and assemble a specific series of products, but also iterate on those products based on consumer feedback and predictive information.
Modular production enables customization
Before we reach a world where humans are largely uninvolved with manufacturing, modular design can help existing factories become more flexible.
Modularity allows the factory to be more streamlined for customization, as opposed to the uniformity that’s traditional for the assembly line. Modularity could come in the form of smaller parts, or modules, that go into a more customizable product. Or it could be equipment, such as swappable end-effectors on robots and machines, allowing for a greater variety of machining.
Presently, mass-production is already refashioning itself to handle consumer demand for greater customization and variety. 90% of automakers in a 2016 BCG survey said they expect a modular line setup will be relevant in final assembly by 2030. Modular equipment will allow more models to come off the same lines.
Startups are capitalizing on the push toward modular parts.
Seed-stage company Vention makes custom industrial equipment on-demand. Choosing from Vention’s modular parts, all a firm needs to do is upload a CAD design of the equipment they want, and then wait 3 days to be sent specialized tooling or robot equipment. Many existing factories have odd jobs that can be done by a simple cobot (collaborative robot) arm or custom machine, and these solutions will gain momentum as factories everywhere search for ways to improve efficiency.
Modular production will impact any sector offering increased product customization. Personalized medicine, for example, is driving demand for smaller and more targeted batches. In pharmaceutical manufacturing, modularity allows processors to produce a variety of products, with faster changeovers.
Robotics automate the once-odd jobs
Industrial robotics are responsible for eroding manufacturing jobs, which have been on the decline for decades. As a report by Bank of America Merrill Lynch explains: “long robots, short humans.”
But the latest wave of robotics seems to be augmenting what a human worker can accomplish.
Cobots (collaborative robots) are programmable through assisted movement. They “learn” by first being moved manually and then copying the movement moving forward. These robots are considered collaborative because they can work alongside humans.
Whether these are truly collaborative or rendering human labor redundant remains to be seen. After a Nissan plant in Tennessee added autonomous guided vehicles, no material handlers were laid off with the increased productivity. European aircraft manufacturer Airbus also uses a mobile robot, which works alongside humans to drill thousands of holes into passenger jets.
While even the best robots still have limitations, economists fear that automation will eventually lead to a drastic restructuring of labor.
Due to rising labor costs worldwide, robotics are presently causing a new wave of re-shoring — the return of manufacturing to the United States.
Manufacturing jobs in the US have been increasing since 2011. 60% of those came from reshoring of jobs previously located in China. And in Q1‘17, North American firms bought 32% more robots year over year.
A majority of US manufacturers, in a survey conducted by BCG, said that lower automation costs have made the US more competitive.
Robotics have become invaluable for monotonous jobs such as packaging, sorting, lifting repeatedly. Cobot manufacturer Universal Robots says some of its robot arms pay for themselves in 195 days on average. As a whole, the category of collaborative robots are priced on average at $24,000 apiece.
We’ve previously identified more than 80 robotics startups, but for heavy-duty machining, significant market share is taken by big industrials players like ABB, Mitsubishi, Fanuc, and Yaskawa.
In the near term, the reprogrammable nature of cobots will allow manufacturing firms to become more customized and work in parallel with existing equipment and employees. On a longer time horizon, however, robotics will be the engine for moving towards “lights-out” manufacturing.
For certain mass-produced items, 3D printing will never beat the economies of scale seen in injection molding. But for smaller runs, fulfillment using additive manufacturing will make sense.
Using metal additive manufacturing for one-third of components, GE made an engine that burns 20% less fuel than previous designs. As of May 2019, GE’s test fleet using this new Catalyst engine has simulated the equivalent of three years of field operation.
Manufacturers will increasingly turn to 3D printing as mass-customization takes off within certain consumer products.
Shoes have become one popular use case to watch. For example, Adidas has partnered with Carbon to mass-print custom athletic shoes. Additionally, other 3D printing services companies like Voxel8 and Wiiv have positioned themselves specifically for the shoe use case.
Just a few years from now, it may be more commonplace to see mass-customized parts in consumer electronics, apparel, and other accessories — all brought to you by 3D printing. Additionally, if rocket-printing startup Relativity Space is any indication, the technology will also be applied to building large-scale industrial print jobs.
Industrial 3D printing is the hottest segment within the broader space, and many startups are aiming to deliver advanced materials that include carbon fiber or other metals with exotic properties.
6. Quality Assurance
As the factory gets digitized, quality assurance will become increasingly embedded in the organization’s codebase. Machine learning-powered data platforms like Fero, Sight Machine, and Uptake, among a host of others, will be able codify lean manufacturing principles into systems’ inner workings.
Computer vision and blockchain technologies are already on the scene, and offer some compelling alternative methods for tracking quality.
In mass production, checking whether every product is to specification is a very dull job that is limited by human fallibility. In contrast, future factories will employ machine vision to scan for imperfections that the human eye might miss.
Venture-backed startups like Instrumental are training AI to spot manufacturing issues. And famed AI researcher Andrew Ng has a manufacturing-focused startup called Landing.ai that is already working with Foxconn, an electronics contract manufacturer. (Below is a view inside Landing.ai’s module for identifying defects.)
Many imperfections in electronics aren’t even visible to the human eye. Being able to instantaneously identify and categorize flaws will automate quality control, making factories more adaptive.
Blockchain will help with recalls
In August 2017, Walmart, Kroger, Nestle, and Unilever, among others, partnered with IBM to use blockchain to improve food safety through enhanced supply chain tracking. Walmart has been working with IBM since 2016, and said that blockchain technology helped reduce the time required to track mango shipments from 7 days to 2.2 seconds.
With 9 other big food suppliers joining the IBM project, including Albertson’s (the second-largest global supermarket by sales) in 2019, the food industry — where collaboration is rare — could also be better aligned for safety recalls.
Similarly, factories employing blockchains or distributed ledgers could be better positioned in the event of recall. In factories where food or automobiles are processed, a single system for managing recalls could more swiftly figure out the origin of faulty parts or contaminated batches, possibly saving lives and money.
Lights-out warehouses may come even faster than lights-out factories.
With the rise of e-commerce, demand for warehouse space has exploded. Last year, the average warehouse ceiling height was up 21% compared to 2001, and spending for new warehouse construction hit a peak in October 2017, with $2.3B spent on construction in that month alone.
Through the last two decades, the average rental area of warehouses has grown by 60% across the US.
Amazon’s historic $775M acquisition of Kiva Systems is said to have set off an arms race among robotics makers. Riding the e-commerce wave and the industry-wide pressure to deliver orders on time, we’ve witnessed an explosion of robotics startups focused on making fulfillment more efficient. Today, Amazon itself has 200,000 robotic units installed at distribution centers worldwide, including 800 complex, large-scale Pegasus robots.
Some startups such as Ready Robotics and Locus have applied the classic robotic arm to package e-commerce orders, though their collaborative nature makes them suited for a number of industrial tasks. We’ve previously looked at industrial robotics companies that could be targets for large corporates.
6 River Systems raised $46M for its treadmill-shaped warehouse cobot, “Chuck,” which helps warehouse employees with daily tasks.
Some of the biggest names in robotics are turning their attention to warehouse logistics too. In April 2019, the Massachusetts-based engineering and robotics design firm Boston Dynamics acquired industrial machine vision startup Kinema Systems as part of its plans to expand into warehouse robotics.
Manufacturers and hardware-focused investors will continue to hunt for the next robotics maker that’s 10x better than the status quo. And the economics of cheaper and more agile robots may mean we’ll see more robots alongside humans in the short term.
AI for scanning
As computer vision melds with enterprise resource planning, fewer people and clipboards will be needed in sorting, scanning, and spotting defects.
Aquifi, for example, uses computer vision inside fixed IIoT and handheld scanners. Machine vision can measure products dimensions, count the number of boxes in a pallet, and inspect the quality of boxes. Presently, this is often done with clipboards, eyeballing, and intermittent scanning.
3D Infotech uses another kind of machine vision technology called Universal Metrology Automation (UMA), which uses blue light or laser scanning to measure surfaces at high speeds.
Vision will be increasingly crucial for IIoT to “abstract away” a real-time picture of what’s happening inside a warehouse.
8. Transport & Supply Chain Management
Once the product is packaged and palletized, getting it out the door efficiently is a daunting task. With thousands of SKU numbers and orders to manage, the complexity can be astounding — and enterprise resource planning (ERP) software has proliferated to handle it.
But there’s still room for IoT and blockchain to get even more granular with real-time supply chains.
Trucking & fleet telematics IoT
In general, there is poor awareness about where items are in real time throughout the supply chain.
The fleet telematics field saw several large exits in recent years, with Verizon acquiring both FleetMatics and Telogis. IoT and software for shipments will only grow more important as supply chains decentralize and get automated.
Farther out, the advent of autonomous trucks could mean that autonomous systems will deliver, depalletize, and charge upon receipt of a Bill of Lading. This will bring greener, more efficient movement, as well as more simplified accounting.
Uber had a highly anticipated autonomous trucking project, but it was shuttered in July 2018 following concerns regarding the founder’s relationship with Waymo.
Kodiak, founded by an ex-Waymo engineer and Otto co-founder, raised $40M in August 2018 for its vision of autonomous trucking. The company has since grown to more two dozen employees.
Peloton Technology, founded in 2011, is working on a platooning model of autonomous trucking. Instead of individual trucks, Peloton is building a system that allows vehicles to communicate with one another, braking and accelerating in concert. Daimler has moved away from platooning, however, as practical issues with the technology (like ensuring multiple trucks want to travel to the same place in the first place) have emerged.
Also in 2018, Starsky Robotics (below) raised nearly $20M from Y Combinator, Sam Altman, and Data Collective, among others, specifically for long-haul trucking.
Daimler first announced it was working on an autonomous big rig in 2015. In January 2019, Daimler revealed its semi-autonomous big rig codenamed Cascadia. Daimler plans to make it available for sale by the end of the year.
As mentioned above, a number of DLT pilots and blockchain startups are trying to put supply chain management software into a distributed ledger.
The willingness to explore these technologies indicates digitization here is long overdue. The highly fragmented nature of supply chains is a fitting use case for decentralized technologies and could be part of a larger trend for eliminating the inefficiencies of global commerce.
Shipping giant Maersk, for example, is working on a joint venture with IBM is aiming to use a blockchain network to help shippers, ports, customs offices, and banks in global supply chains track freight. Maersk’s goal is to replace related paperwork with tamper-resistant digital records, though the partnership has run into problems attracting carriers for the program. Peter Wolf, general manager of CMA CGM, told “Shipping Watch” that only a joint industry standard could succeed — and that a Maersk-IBM program would only ever work for Maersk.
Meanwhile Pemex, the Mexican state-owned petroleum company, is assisting Petroteq in developing oil-specific supply chain management software. The Petroteq project — an enterprise-grade, blockchain-based platform called PetroBLOQ — will enable oil and gas companies to conduct global transactions. In August 2018, Petroteq began working with a blockchain engineering firm called MetzOhanian to develop applications for PetroBLOQ.
In the future, manufacturers will explore decentralized technologies to make their organizations more autonomous and their belongings (coming or going) more digitized in real-time. Blockchain not only has the promise of simplifying SCM, but also could make payments more frictionless.
There are signs that supply chain hype around blockchain may be past its zenith. According to a Gartner manufacturer survey, only 9% of supply chain leaders have invested in blockchain, and only 19% rank it as an important technology for their business, owing mainly to existing blockchain technology’s slow progress towards its promised utility.
Manufacturing is becoming increasingly more efficient, customized, modular, and automated. But factories remain in flux. Manufacturers are known to be slow adopters of technology, and many may resist making new investments. But as digitization becomes the new standard in industry, competitive pressure will escalate the inventive to evolve.
The most powerful levers manufacturers can pull will come in the form of robotics, AI, and basic IoT digitization. Richer data and smart robotics will maximize a factory’s output, while minimizing cost and defects. At the unmanned factory in Dongguan, employing robotics dropped the defect rate from 25% to less than 5%.
Meanwhile, as cutting-edge categories like blockchain and AR are being piloted in industrial settings, manufacturing could eventually be taken to unprecedented levels of frictionless production and worker augmentation.
In the words of Henry Ford: “If you always do what you always did, you’ll always get what you always got.” To reach its full potential, the manufacturing industry will need to continue to embrace new technology.
This report was created with data from CB Insights’ emerging technology insights platform, which offers clarity into emerging tech and new business strategies through tools like:
- Earnings Transcripts Search Engine & Analytics to get an information edge on competitors’ and incumbents’ strategies
- Patent Analytics to see where innovation is happening next
- Company Mosaic Scores to evaluate startup health, based on our National Science Foundation-backed algorithm
- Business Relationships to quickly see a company’s competitors, partners, and more
- Market Sizing Tools to visualize market growth and spot the next big opportunity