How industrial IoT is making steel production smarter

A new project between General Electric and Gerdau, a large steel manufacturer in Brazil, had a simple objective: Improve productivity by reducing equipment failure and related problems. Early indications suggest that goal was achieved with flying colors.

The initial proof of concept project involved monitoring 50 of the company’s assets in October 2015 to identify potential issues with assets before they became serious failures. As a result, the project paid for itself in about none months by preventing unplanned downtime. The pilot project cost $1.5 million and is forecast to generate annual savings of $4.5 million per year. Following the pilot project, the monitoring effort has expanded to cover 600 assets. Gerdau can expect to increase its efficiency in the future if the pilot project is any guide.

[Related: Don’t rush your company into an IoT app platform]

This successful project involved coordinating multiple factors:

• Decide which assets to monitor
• Organize and assess monitoring capability
• Build a model that will offer predictive insights
• Translate data insights into management action

Decide which assets to monitor

Steel manufacturing is an extreme environment defined by significant changes in temperature, vibration and safety risks. And in such an industrial setting, the signal-to-noise factor makes it difficult to decide which assets are worth the effort to monitor. “Selecting the right equipment to monitor is important. In fact, there is a whole methodology around critical equipment analysis,” says Jeremiah Stone, general manager at GE’s APM (Asset Performance Management) Solutions group. This project posed an added challenge for GE “because there was a learning curve to discover the critical assets in steel production,” he says.

GE

Organize and assess monitoring capability

Prior to the pilot project, Gerdau already gathered safety data and monitored temperature and vibration data to an extent. “Installing sensors on equipment wasn’t necessary on this project. That’s an expensive process that is usually only done on older equipment,” says Tracy Ford, Global Implementation Manager at GE Intelligent Platforms.

In addition to technical complexity, installing sensors on older equipment is expensive because the machines have to be shut down during the installation. In GE’s experience with this type of project, there can be 10 to 50 sensors per asset gathering a variety of data points, Ford says.

[Related: Internet of things: Early adopters share 4 key takeaways]

Build a model that will offer predictive insights

When sensors detected a temperature spike on one of Gerdau’s assets, GE was able to recognize this spike as a problem because they had created a mathematical model based on past data. “We like to have a year’s worth of historical data from the customer to create a good model. We then review the data, strip out the data that’s not a good fit and create our model,” Ford says.

In Gerdau’s case, the monitoring effort led directly to bottom line benefits. Through monitoring, GE and Gerdau detected warning signs of a possible equipment break down and thus avoided a 72-hour maintenance effort. These benefits were achieved by leveraging GE’s Industrial Performance and Reliability Center, which performs remote monitoring and analysis.

GE

In the all hype around big data analytics, operational realities in obtaining quality data are sometimes ignored. “Accessing the data was a challenge — it came from multiple sources and different original equipment manufacturers (OEMs),” says Stone. “In this project, the focus was to provide an early indication of possible failure.”

The GE approach to data offers several insights. Start by identifying which assets and processes are worth monitoring. That identification and selection process reduces the amount of data requiring analysis. Then, build a predictive model based on historical data. Finally, look for activities that have predictive value. In steel production, a sudden temperature spike may suggest a problem. In other sectors, there may be different predictive data points to consider.

[Related: If you’re not planning for IoT, you’re already behind]

Translate data insights into management action

Data and reports are helpful but they do not translate into business value unless management takes action. To Gerdau’s credit, company management reviewed the data and found ways to improve their processes. “Our initial catches were not mechanical failures — they were discovering maintenance process improvement opportunities,” Stone says. Rather than simply doing a small amount of maintenance, management changed the maintenance process to reduce the likelihood of a recurrence.

In contrast to financial reporting, where managers may review the books on a monthly basis, Gerdau set up an alert system. By directing high-priority notifications through text messages, Gerdau managers are informed as soon as a potential problem is detected. By scoring and segmenting the alerts by priority, fewer than a dozen alerts are received per month. This means that managers do not become numb to the information. In August, Gerdau received four “priority 1” notifications. In September, there was only one “priority 1” notification.

Next read this: Making IoT magic