Big Data Analytics Improve Manufacturing Performance

May 27, 2015
At the ARC European Forum, Intel and Dow Corning detailed how they have used data to make the right operating decisions and save money.

The theme of the 2015 ARC European Forum in Amsterdam, “The Information Driven Enterprise in a Connected World,” was intended to help develop some clarity over emerging related concepts such as Industrie 4.0 and the Industrial Internet of Things (IIoT). At the forum, we heard many compelling use cases about how applying these concepts can deliver tangible benefits in real-world industrial production across a broad cross-section of industrial sectors. These included presentations about success stories at both Intel and Dow Chemical.

Mitsubishi Electric and Intel presented the results of three IoT pilot applications at one of Intel’s semiconductor fabrication facilities. Intel’s manufacturing equipment generates gigabytes of data per week, per unit, but much of it was not being put to good use. The data includes parameters, error logs, events from machines, and images from vision equipment.

According to the presenters, Intel used Mitsubishi Electric C Language controllers of the Melsec-Q series, which offer robust network connectivity and high computational performance with high availability in potentially harsh environments. Cimsniper data acquisition and processing software was used to selectively transfer process data at rates on the order of mega-bytes per second via a CC-Link IE protocol for storage on a Cloudera Hadoop on-premise, private cloud-based Big Data Analytics Server (BDAS). Revolution R Enterprise from Revolution Analytics was used to analyze the data. The results are transformed into operational intelligence when presented on dashboards accessible through webservers. Not surprisingly, all the equipment used Intel’s high-performance processors.

The company evaluated three test cases. The first aimed to reduce incorrectly rejected units by automated test equipment. The analytics were able to predict 90 percent of potential tester failures to significantly reduce rejection of good units. A second case predicted issues in soldering related to process deviations, reducing equipment downtime and enabling proactive maintenance. A third case concerned image analytics and automating visual inspection of marginal quality units. The image analytics reduced the selection time by a factor of 10 compared with the manual method. Intel published a detailed white paper on the pilots. In the presentation, the consortium reported $9 million of savings during the pilots.

In their compelling presentation, Lloyd Colegrove and his team from the Dow Analytical Technology Center said that the company often underuses, misuses or misinterprets much of its data. Colegrove argued that when put in a much wider context, analyzed automatically, presented attractively, and enabled to act on weak signals, data provides “wisdom” that can guide the company. In this manner, the data can help “justify actions to fix, guide actions to improve, and prescribe actions to make breakthrough changes.”

Dow Chemical partnered with Northwest Analytics to use that company’s Focus EMI solution to take in data from appropriate data sources and present results to all levels in the company for strategic, tactical and real-time purposes. The company went through a culture change.

Colegrove illustrated this with a case in which guessing and post-mortem analysis after a plant trip was replaced by acting upon an analytics-fueled dashboard to be able to make the right operating decisions in time to prevent equipment degradation.

He showed how some plants gradually improved performance, moving from dashboards with many red indicators to those with mostly green indicators. The problem wasn’t that the information wasn’t available before, it just wasn’t either given the right focus or there were insufficient resources to analyze the information. In the next stages of the project, Dow plans to roll out the systems to other business units, continue to build on its knowledge base, expand the use of collaboration, and continue to “develop, partner and dream.”

ARC continues to identify and report upon use cases of how IIoT-enabled Big Data analytics are helping industrial organizations improve manufacturing performance to achieve significant benefits. Readers can check out many of these on our new IIoT/Industrie 4.0 blog.

While the Intel and Mitsubishi IoT pilot did not directly involve the manufacturing execution system (MES) already in place, analyzing real-time production data and presenting these to operators and management for decision support is, in fact, part of the ISA-95 functional scope. Intel, Mitsubishi Electric and Dow Chemical showed that modern architectures, applications and methods used for industrial data analytics open up new opportunities in the operations management realm.

>> Valentijn de Leeuw is vice president of consulting at ARC Advisory Group’s European organization. He has extensive experience in best management practices in process industries, including chemical, polymer, metals, energy, food, pharmaceutical and petroleum manufacturing. His experience includes knowledge of unit processes, simulation and modeling, and business practices using application software designed for manufacturing operations.