The idea of a smart, digitally connected factory is not new. Deloitte describes a smart factory as “a flexible system that can self-optimize performance across a broader network, self-adapt to and learn from new conditions in real or near-real time, and autonomously run entire production processes.” With that definition in mind, the way to get there is fairly straightforward, albeit difficult to actually set into motion. By connecting assets to create data streams, optimizing that data, using those optimizations to find areas to automate, tracking throughout the entire supply chain, making the raw materials as circular as possible, and limiting the carbon footprint, a factory can achieve “smart” status. This article will walk through those steps through the lens of startups.
The first step in making any factory smarter is to connect the assets and start generating data. This may involve adding sensors to production line machinery, installing cameras in key locations, or even getting wearables for workers. The idea is to have a digital read of what is actually going on. Danish pump manufacturer Grundfos is a great example of this. They have added sensors to their production lines that allow them to create digital twins of their machinery. The CTO explained the move to digitization, stating, “When your organization knows earlier when something’s wrong, you are able to provide additional services over your equipment.”
Signal processing companies like Reality AI are helping factories get up and running in the digital space by connecting sensors like accelerometers, vibration sensors, and voltage sensors, and funneling that data into their platform. They use their expertise to collect the right data in the right way because, in their own words, “project teams [tend to] collect data for their first machine learning projects without fully understanding or thinking through the implications for model construction, training and validation. As a result, they don’t get the results they want, don’t get the results they could have gotten.” Getting clean, useful data is a crucial first step in digitization.
Perhaps even more important than collecting data on the factory is using it to optimize workflows. If the first step is collecting data on assets and their performance, the second step is visualizing that data in meaningful ways so that the end user can start to look for patterns. For example, finding defects in the final product while simultaneously analyzing the signal from machinery can help correlate issues back to the manufacturing process and ultimately cut down on errors. On a larger scale, by analyzing data right up to a factory shut down, manufacturers can posthumously go back and find signals to watch for in the future to prevent a potential catastrophe.
One startup that is addressing this step in the journey to a smart factory is Falkonry. Falkonry takes in industrial operations data and applies machine learning algorithms to find patterns in real time. Those patterns can be anything from chemical batch quality to the maintenance needs of machinery. Understanding and even predicting maintenance allows teams to optimize machine utilization and reduce downtime.
Once factory data is being collected and optimized, the way to improve yield even more is by increasing productivity. A New York Times article discussed productivity of facilities as a driving economic force behind technological advances. “Productivity — how much value the economy generates in an average hour of work — gets less public attention than more intuitive economic concepts such as employment and wages, but it may be even more fundamental. Rising productivity — whether through better technology, more educated workers or smarter business strategies — is why people’s economic fortunes, on average, improve over time.” An obvious way to increase productivity through technology is by automating repetitive tasks.
Robots can often carry out these parts of a business significantly faster and more efficiently than humans. For example, startups like Grabit are taking on the once labor-intensive task of pick and place. The electroadhesion head can handle objects as fragile as an egg or as heavy as a fifty pound box.
However, automation does not need to mean a total switch to robotics. The startup Drishti is a perfect example of using automation tools to empower human workers to do their jobs better. Drishti takes videos of humans on assembly lines and converts it into actionable data, offering workflow support for line operators to allow them to perform tasks faster and with fewer errors.
Even if a factory is completely optimized and automated, there is no way for it to move fully into the digital age until the systems around it are digitized. One of the most critical related systems is the supply chain. Goods moving in and out of the factory need to be easily traced in order to accurately predict demand. This type of tracking has a direct feedback loop with the factory automation. When asked about technological enablers to supply chains, the Director of Global Sourcing and Manufacturing at Converse said, “First, data information and analytics. There is so much information — handling it is key to improve visibility and forecasting.”
The easiest way to add a level of traceability into supply chains is to literally put sensors on products and containers to track them all the way through to the customers. Shoof is doing just that. They wirelessly connect assets to the cloud using IOT tags and base stations, allowing for granular tracking in motion.
Circularity of materials within a manufacturing process is not only an environmental cause, it’s also an economic one. Most final products contain only a small percentage of the raw materials involved in the process of making them. By cutting down on the amount of scrap produced in the first place and finding a way to reincorporate the necessary scrap, companies can save huge amounts of money on raw materials.
The aptly named Circulor is helping make strides on this issue by using blockchain, IoT and AI to give raw materials a dynamic identity. This means that the materials can be traced from source all the way through to their incorporation in a product. Just as with connecting machinery, gathering this kind of data is crucial for empowering the circularity of a business.
While a fully lights out factory may be the goal, there are a lot of steps before that to cut back on the amount of electricity and resources a factory is using. According to the Environmental Information Agency, the industrial sector accounts for approximately 40% of total delivered energy use. This is an enormous opportunity for savings.
Crocus Energy is helping factories reduce this energy footprint by bringing transparency to their energy usage. They use machine learning to build models of a facility’s electric distribution system to enable optimization. They have been able to reduce electricity usage in their customers' large facilities by up to 4% using their technology, which accounts for a huge number of kilowatt hours considering the size of the plants they are targeting.