Cognitive automation applications in manufacturing technology
Updated: Mar 25, 2020
Manufacturers of all sizes are slowly deploying analytics and cognitive automation (CA) solutions to parts of their manufacturing processes to improve operations. Cognitive automation is a subset of artificial intelligence (AI) and is able to perform tasks that only humans have previously been able to perform. The addition of cognitive technologies to business processes that are information- or labor-intensive for humans enters the domain of cognitive automation. Cognitive automation is based on software bringing intelligence to information-intensive processes.
Manufacturing was the first major industry to really embrace what we today call cognitive automation. This was with the advent of robotics in the automotive industry in the 1980s, which ultimately did not advance as quickly in other industries with less capital to invest. “Two critical components were cost prohibitive: rapid computing power and the ability to harness a sufficient volume of data,” said Tim Kulp, vice president of Innovation & Strategy, Mind Over Machines. “In the 1990s, processing power started to increase exponentially as costs fell, and today, we have massive amounts of computing power at our fingertips at relatively low cost, and we have digitized almost every piece of equipment in a plant, so we have huge amounts of data to work with.”
Today, it is relatively inexpensive to collect data from machine tools, operators, sensors, and application programming interfaces (APIs) that can pull in information from ERP, CAM, tooling, and maintenance systems.
LEVERAGING THE DATA IN OPERATIONS
The biggest challenge is determining how to strategically use all the data that can be collected. While simply displaying the data can provide information that can improve productivity, processing it with analytics software to interpret it, look for patterns, and to detect anomalies is much more valuable. For example, data from machine monitoring applications coupled with analytics software and CA can provide notifications and predictions of equipment problems, enabling plants to prevent them before they happen.
“Another benefit of collecting data is — even if you are not currently looking at it — you can always look back at it if something fails and look for patterns and anomalies in the data and then apply it to all the other pieces of automation you have. This ties into machine learning because the more data you can feed into algorithms, the smarter they get,” said Mike Cicco, CEO, FANUC.
One might argue that CA is advancing faster than the ability of most companies to fully utilize it. The manufacturing industry is estimated to generate twice as much data as any other industry in the world. Terabytes of data are produced by the typical manufacturing plant every day, and it is doubling about every year. Significantly, it is widely estimated that less than one percent of the data is typically leveraged strategically or even necessary for real business intelligence.
Manufacturers who want to integrate analytics and CA into their manufacturing processes must first strategically determine what processes are most critical to their specific needs and business goals: market growth, profit, quality, ROI, reduced energy consumption, transportation costs, or other company goals.
Best practices in the implementation of CA is to start small and focus on specific operational datasets and results. For example, a specific challenge such as a well-known quality, throughput, or process issue (e.g., machine downtime) in operations should be targeted. Solving a specific problem offers multiple advantages: clearly defined data sources, easily measured success, and easily calculable ROI.
Additionally, “the project should ideally be led by someone in manufacturing engineering or operations management with authority over the operations being analyzed. Their pragmatic view of the goals of continuous improvement and their focus on solving a specific manufacturing challenge will be important in making the project successful,” said Jon Sobel, CEO and co-founder, Sight Machine.
GETTING STARTED WITH ANALYTICS
Sight Machine, a speaker at a recent AMT Manufacturing Technology Council web event, has developed a software platform that uses AI and advanced analytics to enable real-time visibility and actionable insights in a manufacturing enterprise. It can analyze overall equipment effectiveness (OEE) and can provide data, analysis, and comparisons among OEE numbers by machine, shift, or by plant. It can improve throughput by identifying bottlenecks and performance drivers impacting yield as well as determine the root cause of complex plant problems. Once best practices are modeled from the data, they can be applied to all other operations.
“Talking to customers in manufacturing, we learned about common problems — data visibility, real-time understanding of operations, throughput, and improving processes,” said Sobel. “We transform messy raw data into models of useful information.”
The Sight Machine platform was built for discrete and process manufacturing, and data from sensors, PLCs, product data such as serial codes pulled from legacy MES/FIS and ERP solutions, images from cameras, worker IDs, shift codes from scheduling systems, and supplier data can all be part of an analysis.
FogHorn Systems, another speaker at an AMT Manufacturing Technology Council web event, is a leading developer of “edge intelligence” software, uses machine learning and advanced analytics on-premises to enable real-time monitoring and diagnostics, machine performance optimization, and proactive maintenance. Its software can process and enrich any kind, amount, and velocity of data to deliver real-time analytics.
“As a customer of ours likes to say, ‘There is no Six Sigma environment in manufacturing. There is Four to 4.5 Sigma environment, and it’s getting that incremental improvement heading toward Six Sigma where trillions of dollars of value is created,’” said David King, CEO, FogHorn Systems.
Edge computing, which enables data to be processed as it is streamed, is much less expensive for processing huge amounts of data than sending it to the cloud for processing. This enables real-time automated feedback loops in a manufacturing process. For example, FogHorn’s edge intelligence software enables manufacturers to analyze sensor data in real time to identify any values that fall outside of previously defined thresholds and automatically stop the production of defective parts.
As more and more processes become automated and equipment digitized, a fully integrated and visible digital thread will slowly emerge that will connect all equipment and data in a manufacturing process and extend both up and downstream from suppliers to distributors.