agsandrew - Fotolia
Manufacturing, the colossal old guard of the economy, is inching more clearly into the future of expanded automation.
Ultimately, manufacturing and its entire supply chain will reach near-total automation, a paradigm shift as dramatic as that represented by the Internet, experts said. Manufacturing is emerging from a brick and mortar mentality, realizing that long-term sustainability pivots on the ability to exploit machine-learning algorithms by employing predictive analytics.
The simplest and most general application of predictive analytics in manufacturing is preventive maintenance, which anticipates equipment deterioration. Replacing moribund parts before failure prevents unexpected and costly stoppages.
Armed with sensor data, predictive analytics can also inform a manufacturer about ways to more advantageously place equipment and personnel, run assembly lines more smoothly, manage inventory more efficiently and streamline process controls. Superior deployment of resources cut costs.
Bulk of manufacturing in early stages of adopting predictive analytics
Although industries such as finance and insurance are frequently fully analytics-automated, the bulk of manufacturing is wading in the emerging stage of adoption. "Across the manufacturing clients we work with -- no higher than 20% would be called an analytics competitor," Jack Phillips, CEO and co-founder of the International Institute for Analytics, said. "In manufacturing there is still a prominent culture of making decisions based on experience and guts."
Manufacturers continue to rely heavily on charts and spreadsheets. "Even in 2015 we're finding that very large, multinational corporations still have manual processes such as Excel and pie charts to make forecasts," Mike Hitmar, product marketing manager, manufacturing & supply chain at SAS, said.
Still, the pockets of manufacturing that do embrace predictive analytics are setting a strong tone, largely due to the prevailing power of what's commonly called the Internet of Things (IoT) -- the interconnecting of a web of devices that send data to frameworks in real time.
Predictive analytics in manufacturing evolves from buzz to action
With flawless predictive analytics and without limitations of storage, connectivity and computing power, all machines, theoretically, could automatically act on analysis of constant streams of data. In other words, the interconnected machines would self-correct, hyper-accelerating innovation to the tune of casting forth the next industrial revolution.
German automaker BMW is moving in that direction by outfitting test cars with thousands of sensors that send data to the factory whereupon predictive analytics -- supplied by IBM -- offers adjustments that can be made manifest as soon as the next manufacturing stage. IBM business analytics director Erick Brethenoux said that overall technical improvements in recent years have combined to create a far superior product at a lower production cost. "The sensors are smaller, more durable, more capable, more precise," he said. "Also, computing power is stronger and faster because of the ability to parallelize in frameworks like Hadoop."
Ingo Mierswa, CEO and founder of open source platform Rapid Miner cited a current customer (a major European concrete manufacturer) that is using predictive analytics to bolster throughput that recently increased from 70% to 95%. What's more noteworthy, however, is the trajectory of automation. "In 6 to 12 months this predictive analytics process will be completely automated," he said.
Manufacturers are increasingly valuing visibility into an entire operational process in real time. Bryan Tantzen, senior director for IoT at Cisco, said at the World Forum of the Internet of Things in November 2013 that manufacturers sensed the buzz swirling around the IoT but at the same conference a year later -- October 2014 -- the buzz evolved into action. "Six months ago all the major manufacturers I met had initiatives underway to pilot and deploy the Internet of Things capabilities in their plants to drive predictive analytics," he said.
Real-time analysis of big data enables newfound possibilities
The learning curve can be substantial. Converging siloed networks, unifying machines and adopting more flexible compute models represent tall tasks. And although the cloud is the definition of efficiency, it comes with security concerns that have not been thrashed out. Also, manufacturers must decide between spearheading proprietary solutions or opt for the open source route. Proprietary solutions dominate the market and offer valuable services. But the open source way is less expensive and more flexible.
Rapid Miner's Mierswa said the chief advantage of such flexibility is the ability to gain access to the "black box," or the set of algorithms. "It's very important to be able to change the algorithms because every manufacturer has unique ways of doing things," he said. "Perfectly matching algorithms to the needs of a company can create the best potential for success."
Yet the power of real-time analysis of big data is enabling newfound possibilities. Several years ago John Deere realized that the sensors on its tractors and machinery provided data so valuable that it could be leveraged to layer another business model. The sensor data not only informed predictive analytics on optimal ways to operate equipment, it also furnished data on soil and crop conditions and could be used to guide improved planting and harvesting strategies.
General Electric CEO Jeff Immelt has publicly announced that GE is investing heavily in the data gathering and analytical business to provide another service to its customers.
Pratt &Whitney also echoes this trend as it leverages maximum value from the plethora of sensors attached to nearly every machine and device. The sensors' data feeds into predictive analytic models that identify anomalies not otherwise observed. The result is information that yields accurate forecasts on optimal maintenance of aircraft and other equipment.
Providing outcome-based services becoming common
Phillips said the direction of John Deere, GE and Pratt & Whitney will soon attract many followers. "Providing outcome-based services is becoming the common playbook among the 20% of manufacturers that understand the value of data and analytics," he said.
Bill Jacobs, director of product marketing at Revolution Analytics, which supports the open source "R" language, said the integrated supply chains in the automobile, shipping and oil industries are showing the future for manufacturing across the board. "I think the Internet of Things in manufacturing will be ubiquitous in just a few years," he said. "The realities of the supply chain will demand it."
As Forrester analyst Michele Goetz put it: "The holy grail of predictive analytics in manufacturing is the "just in time" supply chain." However, the future of fully automated manufacturing operations and virtually instantaneous processes across the supply chain comes with side baggage. The movement of oceans of real-time data raises a bevy of legal issues. U.S. laws governing the movement of data comprise a disparate collection of state and federal regulations.
In Europe, data movement laws are more regulated. And already Germany has enforced a regulation that inhibits the full potential of the speed of data. IBM, for example, recently facilitated a plan to more efficiently draw energy from the grid by collecting sensor data from appliances used by the German population. The plan was to recommend better times to use electricity to curtail costs and consumption. But the public bridled at the idea of the government knowing too much about their home activities. "We have the ability to mine data in people's homes in Germany every millisecond," IBM's Brethenoux said. "But the law says that we can mine this data only every 15 minutes. This is not bad because the public needs to be protected."
What's the link between RFID and IoT?
What is the Industrial Internet of Things?
Manufacturing intelligence tools gain popularity