Optimization engineers from the OEM or third-party service organization are typically selected for an assignment based on their experience with the specific process. Process generalists don’t have the in-depth knowledge to most efficiently analyze and correct process issues. These process specialists arrive on site equipped with an in-depth equipment and process familiarity that enables them to approach their investigation in the most efficient manner possible and quickly develop appropriate solutions.
Collect Data
Diagnosing the problem requires a series of steps that begins with data collection. The optimization engineers’ objective is to gather as much data as possible. More data generally provides better results, leading to a faster and higher-value process improvement.
“They first determine whether usable data already exists,” says ABB Senior Optimization Engineer Pete Tran. “Most useful data may be stored in a data historian, an archiving tool that captures and stores information for later analysis. Unfortunately, most process owners maximize the amount of data their historian stores by removing much of the detail needed for problem analysis. While the historical data may be of no use, optimization engineers can easily reconfigure the customer’s historian to begin collecting new, uncompressed data.”
Most process automation systems installed in the last 10 to 15 years are OLE for Process Control (OPC)-based. This data communication standard makes it very simple to download process data from any OPC-compliant device or system.
There are still many proprietary process control systems in use, containing possibly useful - but non-standard - data. Experienced optimization engineering firms have converters or translators that enable them to access the data resident in these legacy systems.
“There may also be off line data, such as the results of the regular product testing done in a lab environment,” according to Tim Murphy, Senior Optimization Engineer, ABB Services. “However, this data is typically not very useful. Online data taken directly from the process is captured in a steady stream at a relatively high sample rate of one data measurement every second or five seconds. Lab data, on the other hand, tends to be more along the line of one sample every hour.”
Look for Correlation
While data collection requires special expertise, it is the easy part of the process and within the abilities of most process owners. What they lack is the analysis tools and expertise to turn that data into information that enables analysis. The raw data is nothing but that: masses of numbers. In the hands of skilled optimization experts, those numbers hold the answers to maximizing process performance. Teasing out those answers is a complex undertaking.
“The optimization engineers process the data to help visualize issues, making patterns or trends more apparent,” says Kevin Starr, R&D Manager, ABB Process Automation Services. “The human eye is very good at identifying problems when they are properly visualized. It‘s like a doctor reading an x-ray. There‘s no numerical analysis of the bone to say it‘s broken. Rather, it‘s all visual. The same concept applies for the validation and verification of a pattern in the data.”
Using a variety of mathematical, statistical and analytical tools, optimization engineers search for data points or trends that relate to the occurrence of the problem. One of the most commonly used tools is cross-correlation, based on the search for a certain pattern in the data that relates to a process variable: “When A occurs, B also occurs.”
“The investigation process generally narrows the hundreds or thousands of potential problem points to a very small number of variables, typically one or two,” says Tran. “Even in complex processes where many things can affect quality, it’s possible to narrow the field of candidates to isolate the suspect variables or factors that have the potential to elevate performance and quality. At this point, those variables are still only suspects, innocent until proven guilty through process testing.”
Check back next week for the rest of the “first way” and as always, we look forward to your comments.