One of the most rewarding parts of my job is getting to watch different industries implement new technologies that improve and transform business operations. Manufacturing, in particular, has always captivated my attention in this respect. When I think about how Amazon’s globally connected distribution network has changed in the last decade alone, it’s incredible. From the Internet of Things (IoT) to Artificial Intelligence (AI) and task automation to predictive maintenance technology, the advancements in this space are creating a world of new opportunity.
But this is complicated by that fact that many manufacturers have been around for decades or longer. Some of their equipment was designed before the internet even existed. If replacing this equipment isn’t an option, how do these manufacturers begin their journey to modern manufacturing? The choice of what to embrace and where to start can be daunting.
Ultimately, the reason for adopting any new technology in manufacturing is usually to achieve one or more of the following objectives: produce more, increase safety, or increase quality—and all at a lower cost. The good news is that the most important thing a manufacturer needs to accomplish with any of these objectives is something they already have. It’s something they’ve had since the moment they opened their doors, whether that was yesterday or 100 years ago: data.
In some ways, data is to factories what oxygen is to humans. Oxygen is all around us, but it doesn’t do us any good until we breathe it in. Then we can use it to create energy. The more energy we have, the more we can do. Manufacturers all over the world are sitting on mounds of data that they aren’t “breathing in.”
Truly, the journey to modern manufacturing begins with unlocking this data. The ability to tap into that data is the difference between factories that are “smart” and benefiting from new technology, like automation, machine learning (ML), and AI, and those that are being held back by outdated tools and systems.
Where is this trapped data? That’s where the problem lies for many manufacturers. A lot of this valuable data is sitting in places where it’s hard to collect, compare, and take action on, like in old machinery, isolated systems, spreadsheets, and even paper clipboards.
In an environment with varying legacy protocols and data formats from the last 30+ years, how can one compare important metrics, like productivity, equipment availability, and output quality, from different lines within a factory or different factory sites? The data is stuck in silos. I know this because, at AWS, we’ve helped hundreds of manufacturing companies, like Georgia-Pacific, Volkswagen, and INVISTA, solve the challenge of liberating industrial data to extract insights from it.
Using the right tools
Manufacturers that are struggling to modernize their factories or feel overwhelmed about all the different investments they could make in new technology should focus on getting to the data. Collect data from on-premises equipment, historian databases, and IoT sensors and move it to the cloud. Then it can be organized, analyzed, visualized, and, eventually, used to do more advanced things, like train an ML model that can help predict when machinery will need to be serviced, thereby avoiding unplanned downtime for operations. While that might sound easier said than done, there are a variety of solutions that can help.
Let’s take a look at the journey to modern manufacturing and some of the tools at AWS that can help along the way.
One solution that’s become very important at AWS in simplifying the ingestion of data into the cloud is AWS IoT SiteWise, a managed service that makes it easy to collect, store, organize, and monitor data from industrial equipment at scale.
Figure 1: AWS IoT SiteWise process flow
AWS IoT SiteWise includes edge gateway software that automates the process of securely connecting to on-premises equipment, collecting and organizing industrial equipment data, and sending the data to the cloud. Manufacturers run this gateway software on popular third-party industrial gateways to read data using OPC Unified Architecture (OPC-UA). This is an open interoperability standard for secure and reliable exchange of data within industrial automation developed by the OPC Foundation.
By using standards like OPC-UA, industrial customers have a consistent method for interfacing with many types of industrial equipment, like SCADA systems, PLCs, and historians on the factory floor. With this information, manufacturers can compare machine 1 to machine 2 (which may have been bought 15 years apart), production line 1 to production line 2, and even factory 1 to factory 2.
AWS IoT SiteWise gateway software is provided as a pre-packaged connector that runs on AWS IoT Greengrass, which extends AWS to edge devices so that they can act locally on the data they generate while still using the cloud for management, analytics, and durable storage. There are millions of connected devices in the manufacturing world collecting important data every minute, like the environmental, process, and vibration data from a particular machine.
AWS IoT SiteWise also provides interfaces for collecting data from modern industrial applications through MQ Telemetry Transport (MQTT) messages or Application Programming Interface (APIs).
To make the data useful, you have to give it context. With AWS IoT SiteWise, customers model their industrial equipment, processes, and facilities by adding context (like equipment type and facility location) to the collected data and creating hierarchies to represent relationships. Then they define common industrial performance metrics, like Overall Equipment Effectiveness (OEE) and uptime, on top of the data using the AWS IoT SiteWise built-in library of mathematical functions.
As data is ingested into the cloud, AWS IoT SiteWise automatically computes the metrics at the interval defined by the customer (like “report uptime every hour”). All uploaded data and computed metrics are stored in a fully managed time series database, which automatically scales with the customer’s data usage and storage. This type of data store is uniquely designed to store and retrieve time-stamped data with low latency, making it significantly easier for customers to analyze equipment performance over time.
For example, Bayer Crop Science, one of the largest agricultural companies in the world, is constantly striving to optimize yield in their crop fields and reduce waste in their production plants. A major challenge they have faced in doing this is making all the data they collected useful because it was stuck in siloes. Now, using AWS IoT SiteWise across multiple corn production plants, they can combine and analyze this data in a meaningful way, like to measure the OEE of their machinery in near real time to identify production inefficiencies.
Having the right tools to collect, organize, and create metrics on incoming data without writing any code creates huge savings in time and effort for any developer. But the data and insights gathered also need to be easily shared and visualized by industrial end users, such as process engineers and operators, who will use the information to identify and apply corrective actions and process improvements.
From within the AWS IoT SiteWise console, customers can create no-code, fully managed web applications in minutes. There, they can display equipment data and computed metrics in near real time and compare and analyze historical performance across equipment or facilities. End users can access the web applications from a browser on any web-enabled desktop, tablet, or phone and sign-in with their corporate credentials through a single sign-on (SSO) experience. Customers can create one or more web applications to easily share access to industrial data with any team in their organization to spot anomalies. This helps manufacturers reduce waste, make faster decisions, and optimize their plant performance.
Figure 2: AWS IoT SiteWise end user web application
Machine learning and more
The next step in the journey to modern manufacturing is to start using ML for predictive maintenance. Many manufacturers rely on people to perform routine diagnostic duties and preventative maintenance on fixed schedules. ML can provide a more reliable approach to preventive maintenance. ML models can help predict the likelihood of asset failure using sensor data and optimize schedules for maintenance procedures. This predictive maintenance can help lower maintenance costs and reduce unscheduled downtime.
For example, Georgia-Pacific is using ML to optimize paper towel manufacturing. The Georgia-Pacific ML model can predict—based on the quality of a parent paper towel roll—precisely how fast converting lines should run to avoid tearing and to maximize production while still maintaining the best quality. By reducing paper tears, Georgia-Pacific has increased profits by millions of dollars for just one production line. And there are at least 150 more lines that could benefit from these optimized processes.
Predictive maintenance requires ML models that are trained on large amounts of data. This is why data ingestion and data management are critical first steps to take before expanding into ML. It’s also why many companies choose to create a data lake, a centralized repository where organizations can store all their structured and unstructured data at any scale. A data lake is a powerful foundation for ML and AI (artificial intelligence) because ML and AI thrive on large, diverse datasets.
An important tool many of our manufacturing customers use for ML, including Georgia-Pacific, is Amazon SageMaker, a fully managed service that removes the heavy lifting from each step of the machine learning process to develop high quality models.
The history of manufacturing is a timeline of technology innovation. Since the industrial revolution, manufacturers have been developing new approaches to increasing quality, speed, safety, and efficiency while lowering costs and waste. Now, as manufacturing innovation has moved to the cloud, it's exciting to see the possibilities available to manufacturing customers of all sizes and in every industry.
At AWS, we believe no manufacturer should be left behind. We provide a path that can help manufacturers take advantage of new technology and business opportunities, no matter where they are in their journey. I look forward to seeing what our manufacturing customers will develop next.