| From data to decisions: Turning information into intelligence By Jack Smith, Managing Editor, AppliedAutomation
Collecting data for the sake of collecting data is inefficient and ineffective. Data should be collected for the valuable insight it provides into the manufacturing process and for the higher level business decisions it could support.
Transforming data into information
“Many manufacturers don't realize what data collection really is, and how collected data becomes usable information,” said Ron Iannacone, president of Factory Intelligence Network, West Berlin, NJ. “Data collection has become a catch-all phrase, covering everything from logging machine temperatures and pressures by hand to sophisticated systems that display and record real-time data from hundreds of data points to control systems equipped with data collection capabilities that store data in multiple, disparate factory and enterprise systems.”
Data means different things to different people. The vice-president in the corner office at corporate may be more interested in overall run rate information, while the manufacturing manager may want that same data plus application-specific data. David Crump, marketing communications manager at Opto 22, Temecula, CA said that 'useful' is in the eye of the beholder. “You have to identify what data is important, who it's important to and then get it to them.”
In addition to facilitating on-the-fly robot tool changes at the DaimlerChrysler assembly plant in Belvidere, IL, the DeviceNet connections in the photo allow systems data to be collected, analyzed and stored in real time.
|
Data must be converted into information to be useful. “Information is what you want; data is what you get,” said Richard Theron, marketing services manager at ProSoft Technology, Bakersfield, CA. “The Internet contains a vast amount of data, but often, very little information. The process of gathering useful information begins by understanding what data is required and how it should be viewed.”
“The challenge for manufacturers is that they lack the time or the tools to interpret the data into meaningful information,” said Dennis Cocco, chief product strategist and founder of Activplant, London, ON. “What the plant truly desires is the ability to understand what data to focus on.
“For example, data would be a collection of temperature samples taken every 10 seconds from an oven that is heat-treating parts. Information would be the ability to know whether – based on what part type is in the oven and the specification for heat-treating temperatures – the oven temperatures are correct for this part, or they are out of spec.”
“Making use of context is the start of turning data into information,” said Roy Kok, senior product manager, HMI/SCADA at GE Fanuc Automation, Foxboro, MA. “Context involves location, timing and an understanding of cause and effect. Context is all about being able to relate data to each other, and to happenings in production. The generation of context – through automated analytics – is the key to fast retrieval times. There are many relationships that can be stored along with the data making further analysis more efficient.”
“Data without context is basically noise,” agreed Gregg Le Blanc, product & industry marketing director at OSIsoft, San Leandro, CA. “When you wrap data in context – whether it be metadata, related information, analytics or some sort of other relevant information – raw plant data becomes a powerful tool for decision making.”
“All data must be collected within the context of what it means,” Cocco said. “In the (heat-treating) example, the context would be that we understand not only what the temperature of the oven is, but that we also know what parts are in the oven and how long they have been there.”
“If context is available, you can limit your recalls to exactly the parts that are in question,” explained Kok. “If context is available, you know exactly what raw materials were applied to what production run and what operators were involved. Context can be simple: this value is in this plant, this area, this line, this piece of equipment; or it can be as complex as: this flow affects this tank temperature, which affects this cure time, which affects this elasticity and finally the friction of this tire.”
What to monitor
What should be monitored to provide insight into business decisions? “This is a difficult question to answer,” said Cocco. “Some may argue that it is only necessary to focus on the key and primary processes in the plant. However, we have witnessed many cases where the secondary or less important processes end up being the primary constraints to production throughput.”
“If you are making a decision as to whether to measure one variable versus another, and they're both already instrumented, then no matter what you choose, you'll be missing part of the picture,” added Le Blanc. “The questions you may ask in the future may involve data you may not have anticipated measuring today.
“In today's economy, anything that involves energy usage should be scrutinized. It will render the biggest return, in general. If you have key active ingredients or materials that are either expensive or limited in supply, you should keep a close eye on those as well.”
“If you're not measuring, you're just practicing,” said Ward Komorowski, P.E., director of facilities and building services at Johnson Controls, Milwaukee, WI. “If you're not keeping score, you're just practicing. The ability to see what's consuming what in your process environment: that's the scorecard. Once you see that information, then you can improve it.”
Focus on desired outcome
Making use of information involves defining and leveraging Key Performance Indicators. Theron suggests that the best practices for accomplishing this are to:
- Define the KPIs and stick to these definitions
- Identify the KPIs that will have the best impact
- Monitor whatever leads to ROI.
Focusing on the desired outcome directly affects the methods used. “In general, we have seen that control information (temperatures, counts, faults, alarms, test results) can generate volumes of data that is irrelevant to making business decisions,” Cocco said. “Starting with the key business drivers and KPIs and working back to the data needed to support those KPIs will guarantee relevance of the information to making the right business decisions.”
SEW Eurodrive is automating its entire operation to become more competitive. In addition to collecting data, which becomes useful information for use throughout the enterprise, an advance control system enables the fully-automated, lights-out heat-treating process at its Lyman, SC facility to be run by only one person.
|
But collecting all data on everything without relating it to a context can be almost as ineffective as not collecting enough data. Finding exactly what supports KPIs within raw data that is not tied to a real business context may be virtually impossible; some companies just give up.
“In our experience, the best implementations are the ones that start with the business drivers” said Cocco. “The next step is to understand what KPIs are needed to support the business drivers and the underlying supporting data to understand how to impact those KPIs. Far too often, we see companies take the approach of. 'let's collect all the data and then we'll determine what kind of analysis we can perform on the data later.' This is a dangerous approach because it increases cost and complexity, and generally misses the key business drivers.”
“Business decisions are based on whatever information is available from the manufacturing floor,” Le Blanc explained. “If that data is stale, the business might not trust or be able to make business commitments to customers. Having the data available in near-real-time is important. The business will want access to key variables to do basic trading. These usually include inventory, production rates or quality information.
“More and more businesses will want to go beyond this and look at equipment availability, the capabilities of different sites to manufacture key products that are in demand out in the marketplace and also past performance or repetitive incurred costs. These allow the business to be run in new ways to help innovate beyond the competition.”
Interoperability
According to Cocco, about 90% of the data is collected directly from control sources such as PLCs or DCSs. The remaining data can come from external devices such as bar code scanners or manual data entry.
Whether data is collected from PLCs, DCSs or manual entry, converting it into useful information can provide valuable insight into manufacturing processes and support higher level business decisions.
|
“Programmable automation controllers and I/O systems perform control functions and also provide data acquisition and delivery to enterprise apps and databases,” Crump explained. “Use open, non-proprietary protocols when developing your control hardware. This makes data collection and communication much easier.
“You should deploy systems that have the ability to communicate effectively across the plant floor and beyond. Connectivity shouldn't just be to a local HMI or PC; it should extend to databases and applications throughout the enterprise, if needed. That way, relevant, appropriate data can go into supply chain apps, MES apps, ERP apps or wherever.”
“In today's plants – whether discrete, hybrid or process – automation systems include a mix of solutions installed at different times from different vendors,” said Kok. “A successful philosophy is one of 'Embrace and Extend' existing legacy systems – layering value on top of them and acquiring data from them.”
Open architecture is very important within a production environment. OPC enables multiple vendors to work together nicely within the same enterprise. As new technologies emerge, take advantage of them by applying them to the processes. |
