Manufacturing Intelligence: Using the right tools to make the right decisions

I f you don’t have the data, you don’t know how you can improve,” notes Igor Valuyev, who has served as New Belgium Brewery Company’s chief electrical and automation engineer since 2000. “You don’t know what you don’t know.” (FE, Sept. 2010)

Jump to:

What is EMI?

What data constitutes manufacturing intelligence?

“Before we implemented this system, operators could only see how many ‘bad apples’ were produced,” says Jon Riechert, Hillshire Brands corporate senior project engineer. “Now operators can see weight fluctuations along the line, the exact giveaways on finished chubs, machine downtime, OEE and even meat temperature.”

Most likely you have an ERP system in place. And you may have a process and/or packaging control system. But do you have a manufacturing intelligence (also known as enterprise manufacturing intelligence) system that can look at the data in your ERP, process control and packaging systems, and let you know the real state of your manufacturing operation at any given moment? The engineers at both of the above plants put EMI systems in place and have been rewarded for their efforts: Of the hundreds of measurements at NBBC, 20 key variables allow operators to focus on quality, better process sequencing and peak load shedding. Hillshire Brands has eliminated 100,000 pounds of giveaway sausage every year, increasing true yield and improving quality.

What is EMI?

Enterprise Manufacturing Intelligence software is a crucial tool organizations use to gain immediate, actionable intelligence from both shop floor automation and enterprise business systems, according to LNS Research. “EMI software provides the ability to connect, federate, aggregate and contextualize data from multiple sources into useful and timely information,” explains Mark Davidson, LNS principal research analyst. EMI provides the appropriate alerts, dashboards and displays to support a wide range of roles and responsibilities across a manufacturing/production organization. The dashboards usually support drill-down navigation to uncover additional levels of detail and to perform root-cause analysis.

EMI software typically includes or supports add-on analytical tools to enable “power users” to uncover interrelationships across data sets that provide new insights for making improvements or taking corrective actions. A comprehensive EMI solution goes beyond plant floor-generated information and can pull in business systems information to correlate specific customer, product line or cost information along with production, energy and logistics information—making it a powerful manufacturing plant and business tool.

Many manufacturers are already seeing the benefits of EMI systems. According to LNS Research’s Manufacturing Operations Management survey, 61 percent of manufacturing companies across a broad range of industries either have in place, or plan over the next year to have, manufacturing business performance dashboards. Companies that are not on this path will be at a competitive disadvantage, the study states.

“Manufacturing intelligence is a method of providing detailed dimensions to flat process data,” says Maryanne Steidinger, director of software product marketing, Invensys. “Manufacturing intelligence comprises several facets: contextualization (bringing discrete elements into relationships with one another); persistence (creating and sustaining those relationships so they can be updated with real-time event data); reporting and analysis via interactive dashboards; multi-source [inputs] (all plant data residing in a database must be included for relevancy); and most importantly, a plant model, in which you can create the relationships and details on how you’d like the data to be displayed, structured, updated and presented.”

For a food processor, there are both event-driven and transactional events that occur during the course of processing materials, says Steidinger. For example, traditional “data historian” events, such as temperature, time and speeds, are collected in millisecond increments. Transactional data is also created, which is typically managed in a manufacturing execution system: work order/bill of materials management, material consumption, labor, quality, overall equipment effectiveness, etc. Transactional data has multiple dimensions and adds context and relevancy to the historian data. “So by using a manufacturing intelligence product, you can have not only an alarm, but find the context of that alarm: Was there a line down? Did you have untrained workers on the line? Was there a material mismatch or outage? It helps the food processor better understand the cause-and-effect relationships that occur during the course of manufacturing,” adds Steidinger.

“Food and beverage processors realize each and every process provides a vital link to real-time data for the enterprise as a whole,” says Melissa Topp, Iconics director of worldwide marketing. This real-time data is especially critical with the move toward real-time performance management that uses dynamic performance targets to drive an enterprise to its optimum potential and competitive advantage, adds Topp. Manufacturing intelligence provides a common basis for coordinating production information from the plant floor and enterprise systems as well as the entire supply chain.

“Connectivity of each and every process is essential for an enterprise to achieve operational excellence,” says Topp. “Fact-based analysis drives consensus and improvement, which are impossible without real-time connectivity of both enterprise systems and the plant floor.”

Real-time operational intelligence systems turn data and information into manufacturing intelligence, which can be acted upon within an equipment and personnel model to provide “the right information to the right person at the right time,” says Katie Moore, GE Intelligent Platforms global industry manager—food and beverage. “It’s all about the ‘push,’ not the ‘pull.’ That is true manufacturing intelligence.”

What data constitutes manufacturing intelligence?

Real-time data capture, alerts and actionable intelligence are essential for cost control and performance improvement for food and beverage processors—that is, knowing what the machine or production process is doing right now, at this very moment, says Diane Murray, Epicor Software Corp. manager, product marketing. Processors must be able to capture production, equipment, quality and maintenance events in real time. Beyond these, operators must be able to add depth and dimension on the spot including downtime causes, changeover events and observational reasons for scrap, yield and quality loss.

From there, it’s all about feeding the information to each person in a way that makes sense for his or her role, such as shift scores for line professionals and on-the-fly, slice-and-dice analysis for operations stakeholders, adds Murray. Certain key performance indicators beyond OEE may include the monitoring of labor efficiency, equipment energy usage per shift or SKU, yield standards and cost standards. The KPIs used should drive toward business objectives and align with corporate improvement goals and objectives.

Depending on the type of operation and business, the KPIs to be tracked will necessarily vary. LNS Research’s Davidson divides them into five areas: business/financial, quality, efficiency, production and safety.

For business/financial KPIs: Cost per unit, energy costs and return on assets are important metrics to monitor, given the small margins at which most processors operate.
For quality KPIs: Customer rejects/returns and in-process quality/yield are important to brand integrity.
For efficiency KPIs: OEE and production changeover times are important to meeting the flexibility needs of producing multiple product variations in both processing and packaging areas.
For production KPIs: Capacity, cycle time and schedule attainment/on-time delivery information helps keep producers on track in adapting to dynamic fluctuations in supply and demand.
For safety KPIs: WIP inventory aging, materials and product genealogy, and tracking and traceability are essential to ensuring food safety and compliance.

While these five groupings include the major areas of concern, there are other ways of grouping KPIs, some of which can hone in on production. David Picciotto, Siemens US MES industry management process industries, defines several relevant KPIs, which fall mostly into the previously mentioned efficiency and production KPIs.

Asset-driven KPIs: For instance, throughput of overall equipment effectiveness, which is a combination of quality, performance and availability rates.
Schedule/planning-driven KPIs: For example, conformance to schedule. The KPI calculation is a combination of actual order execution time vs. the planned execution time. This information can be used for campaigns, production orders or batches, providing increased granularity if necessary.
Production-driven KPIs: These include yield-driven (planned vs. actual KPIs to compare expected produced quality vs. actual); scrap-driven (to monitor waste and keep quality under control); and key operational indicators-driven (to monitor a particular process value that has direct correlation to every other KPI).

“From our experience, asset-driven KPIs are particularly interesting for food and beverage processors, where filling and packaging lines are important parts of production,” says Picciotto. Where continuous and batch processing are predominant, just a small portion of asset-driven KPIs, mostly related to downtime management, is typically used.

Obviously food safety-related KPIs are important, but the biggest push for a targeted KPI is OEE, according to Mike Pantaleano, Rockwell regional manager, information software. “OEE is really important in food and beverage, and especially in beverages where there is a large concern for bottling lines. Am I getting the right utilization out of the plant? Can I make as much as what the machines are spec’d for? Can I do more? OEE has its roots in discrete applications and provides the ability to take a look at availability, performance and quality.”

Pantaleano considers OEE as just a grade, a number that doesn’t show all the factors that went into calculating the final score. So, for example, if a line has an OEE of 67 percent, that’s a snapshot in time. More important is whether the score is trending downward or upward. For instance, if a score is trending downward, the processor might suspect an issue with performance or availability.

Where to start

Manufacturing intelligence strategy improves sausage yield by 100,000 pounds

The Hillshire Brands facility in Newbern, TN produces more than 150 million pounds of sausage annually, with the Jimmy Dean brand in one-pound chubs accounting for more than two-thirds of the total production. The facility employs 11 machines to grind, season, fill and package chubs of more than 300 different SKUs. Since the product is priced per package, any meat beyond one pound is a giveaway. At the end of the line, a checkweigher assesses packages for weights within a specified window. Off-weight packages are rejected, and the meat is either put back into the system or discarded.

Jon Riechert, Hillshire Brands corporate senior project engineer, had already implemented manufacturing intelligence strategies and information systems at plants in Florence, AL and Kansas City, KS. After meeting with the engineers in Newbern, he brought in a team from Grantek Systems Integration to implement a manufacturing intelligence strategy.

Grantek worked with Riechert and his team to implement a manufacturing intelligence solution based on Rockwell Automation’s FactoryTalk software suite. The processor already had Rockwell Logix control platforms at all its locations, so it was a matter of installing the right software to track the 11 chub-line machines. FactoryTalk VantagePoint was used to integrate information from multiple data sources: controllers, the historian and SQL server.

Newbern plant engineers worked with Grantek to design system dashboards to help operators spot problems. “The Newbern team had full control over what they wanted to see on the VantagePoint dashboards,” says Riechert. “I’ve worked with this software enough to understand there are almost infinite possibilities of what it can show visually.

“Before we implemented this system, operators could only see how many ‘bad apples’ were produced,” continues Riechert. “Now operators can see weight fluctuations along the line, the exact giveaways on finished chubs, machine downtime, OEE and even meat temperature.”

Armed with this information, Newbern engineers are close to a yield improvement of 0.10 percent for a savings of 105,000 chubs per year and hundreds of thousands of dollars in the cost of goods sold and justifying the ROI on the project within six weeks of implementing the new system. The processor expects the new system will help reduce giveaways by another several tenths of a percent, increasing yield by up to 0.50 percent—or more than half a million pounds of sausage.

Many processors have some form of accounting or ERP software, but some may have little or no software to support the monitoring of processing and/or packaging operations. ERP software can provide a lot of business information. But unless a processor receives data from a shop floor system and compares it against business needs, it is, at best, operating with sketchy shop floor information gleaned from clipboards and Excel files—information that’s a day late and a dollar short—or maybe tens of thousands of dollars short. Unfortunately, a system like this doesn’t support continuous improvement and/or quality and will fall short when it’s time to produce specific data to satisfy a USDA or FDA recall.

Where to start? It’s time to automate at the shop floor level. LNS Research sees automated data collection as the first logical step toward an effective performance management and EMI strategy. This often requires additional instrumentation and data acquisition hardware and software, but may not necessitate a high degree of automation in and of itself. Barcode scanners, wireless scanners, wireless sensors, RFID tags and mobile software applications are all being used to support automated data collection.

According to LNS Research’s survey, 28 percent of companies already have data historians in place to automatically and reliably collect manufacturing/production data; 24 percent are planning to implement historians next year. “Clearly, companies are seeing the benefits of automating data collection processes and are getting away from time-consuming and error-prone manual data collection wherever possible,” says Davidson.

Making use of historian technology, GE’s four-step process, called the Operational Excellence Journey, is designed to drive manufacturing process stability and repeatability. Plant automation tends to be an assortment of old and new, disparate or point solutions that provides isolated value, says GE’s Moore. She describes the journey as follows: The first step is to connect the disparate systems together for real-time visibility and control. The second step is to add context to the data from both a plant asset and role perspective to drive production KPIs. The third step is to add more advanced predictive analytics across multiple areas of the plant, and finally, the fourth step is to integrate that data with the existing enterprise systems such as ERP.

It is critical that the earlier steps are completed before enterprise integration since each step adapts the raw high-speed data (e.g., temperatures and pressures) from the machines and translates it into a format and frequency that make sense for an ERP system to use. Depending on the type of food and beverage manufacturing, the levels of existing automation vary dramatically. Consequently, each manufacturer must identify the areas of highest impact in its process and ensure the EMI solution it chooses provides a pathway to gradually gain incremental value and has an open, layered platform on which to build and connect existing systems and infrastructure.

Traditionally, data historians collect data directly from automation equipment and processes, such as PLCs, sensor data, temperature recorders, etc. At the data historian’s output, data moves to analytical and quality systems, so operators and process engineers can look back into processes that happened an hour ago, yesterday, last week or last year. As manufacturing intelligence software finds its way into plants, infrastructure follows, including hardware, monitors, networks, etc.

“But there’s also a movement to use the cloud,” says Steidinger, “meaning applications can be hosted by someone else and simply used by the processor without the need for on-site computing infrastructure.” SmartGlance, for example, is an Invensys cloud-based reporting and analysis application that can be hosted outside the processor’s site or onsite. This manufacturing intelligence software connects to plant historians, SQL databases and Excel spreadsheets.

If a processor has a real-time process control or packaging system in place, one question that might arise is: From where should data for an OEE KPI calculation be extracted—the ERP system or closer to the automation system? “OEE is a great example of where I wouldn’t rely on the ERP because you’re trying to get information in real time,” says Brandon Henning, Rockwell Automation MES product manager. “You’re getting information directly off the controllers, and you’re trying to use that information right within the operation. Trying to pass it up to an ERP system and have it do the analytics is not an effective use of either the OEE [shop floor] or ERP solution.”

Displaying EMI data

Displays come in all sizes, from four to 60 inches or more. While some processors believe a 19-inch monitor mounted on a floor stand is sufficient for operators functioning in a single zone, their bosses and their bosses’ bosses may no longer want to be confined to one location, whether it’s the plant floor or an office. Indeed, wireless smartphones, tablets and industrial laptops can operate wherever there’s a signal.

One major improvement to the display is not so much technical as it is philosophical, and it has to do with distraction. Some years ago, when distributed control systems began to have full-color displays on monitors larger than 17 inches, it became fashionable to crowd as much of the process as possible on a single screen, along with an alarm system that flashed red every time a process parameter was slightly above or below set point. In many cases, operators would receive one alarm, followed by an avalanche of subsequent alarms. While the process may have been far from going out of control, the operator was so distracted it was almost impossible for him or her to react rationally.

“There is a lot of discussion on ‘situational awareness’ now, meaning essentially that you remove all extraneous elements from a display to show or indicate only those items that are out of spec,” says Steidinger. “It is a much flatter and less graphic-driven mode of management and a huge departure from those graphic-rich HMI screens of the past. The goal of situational awareness is not pretty pictures: It’s to show immediately and effectively changes in the process that need immediate attention.”

Considerations for plant data and who uses it are important. Siemens’ Picciotto lists four important concepts in data types:

• Sources: Automation and operator input are two of the many options. Data may come from quality systems, ERP, a scheduling tool, process optimization software, etc.

• Destination: What user or system requires information?

• Context: What information can be provided to enhance the data and better describe the current situation?

• Actions: Which actions need to be executed when a specific event arises, and which systems/users need to be notified and/or take some actions?

Today, displays and the data they present are set up specifically for the person who uses them. “You need to be flexible enough [to know] that what’s important to a VP of operations might not be as important to a shift supervisor,” says Rockwell’s Pantaleano. Role-based dashboards provide the right amount of information to the user. So, while the VP checks production capability, this information would be useless to the maintenance engineer who has to roll up his sleeves and fix a machine problem before the VP sees a change in the KPI displayed on his smartphone.

“This new wave of monitoring on the go is already becoming the standard way of operating,” says Iconics’ Topp. “Having the capability to monitor your factory and review real-time data is a game changer. Now, lead operators can receive alerts anytime, anywhere on their personal mobile devices, get instant updates and drill down to pinpoint the exact source of trouble in their operation. This allows for better, faster decision-making at critical moments, whereas before, an operator offsite would have to rush to the control room or specific piece of equipment to understand what was going on or rely on relayed information from those at the facility.”

Are we connected?

Manufacturing intelligence systems would be dead in the water if all constituent hardware and software couldn’t communicate. There was a time when systems didn’t connect well because computer operating systems (UNIX, Windows, Mac, OS/2, Linux, VMS, etc.) used proprietary inter-process communications methods, and proprietary wired industrial networks weren’t much easier to connect. Fortunately, most operating systems that survived now have connectivity methods that reach out to other systems, and TCP/IP network architectures and web-based objects and agents bring systems together without all the integration time needed a decade or more ago.

MetricStream, a web-based, end-to-end quality control and quality management system, allows processors to collaborate with their partners, gain a real-time view into their quality data and enable issue-tracking for a closed-loop compliance process, according to Sonal Sinha, associate vice president of industry solutions, MetricStream. The software provides agent-based interfaces called infolets, which can integrate external systems through flat files, messaging interfaces and web services application programming interfaces.

“The typical production environment holds automation and systems from multiple vendors, with varied adherence to standards like OPC-UA, ISA-95 and/or ISA-88,” says GE’s Moore. “The advent of integration tools that bring a service-oriented architecture to the factory floor has reduced the impact of this [integration] problem. For example, GE Intelligent Platforms software solutions are based on the Proficy SOA platform, which enables disparate data sources to be mapped through ‘virtual’ ISA-95- or ISA-88-based data models. This capability eliminates the need to develop multiple integration points from factory systems to a single master quality system as an intermediate step to sharing data.”

“Plant data connectivity sources may include DCS/PLC, historians, SCADA/HMI/LIMS and SPC quality systems that can be integrated through OPC, SQL or ODBC connectors,” adds Epicor’s Murray. Leveraging standard interfaces like OPC/ODBC is an effective way to minimize integration efforts. “Epicor works to minimize or eliminate the burden on IT resources. That means knowing how to get a signal from any kind of machine or monitor a production line at any point. It also means having a robust data exchange module to send and receive data to other kinds of plant or enterprise systems,” continues Murray.

While all the tools are in place for integration, it’s incumbent upon a processor to understand what the end result is since this will determine the types of information needed/shared between applications, according to Invensys’ Steidinger. “We have seen projects where the different information flows are mapped, showing information coming from the business system to the MES and vice versa. In this way, the processor and integrator/service provider can agree upon the information required, e.g., work order downloads, material consumed uploaded and other details, such as determinism (did the transfer get there?), queuing (cached just in case of a break in the interface), etc.”

With all this connectivity, EMI systems can make sense out of your manufacturing data and let you know where to make improvements. These systems often pay for themselves in a short period of time by reducing giveaways and scrap, spotting bottlenecks in production and packaging, and improving yields and efficiency. Continuing to practice the “same old” will only perpetuate the same old, and who can afford that today?