JUL-AUG 2018

Issue link:

Contents of this Issue


Page 18 of 61

INTECH JULY/AUGUST 2018 19 PROCESS AUTOMATION torian recording purposes. The creation and management of large sets of if-then rules for data vali- dation as well as other applications (e.g., real-time process diagnostics, intelligent alarming) is a strength of re al-time expert systems. However, most commercial automation systems can easily handle the use of small to medium if-then-else rule sets. Although mathematical equations are the basis of conventional filters, human logic statements can be pow - erful additional ways to "clean up" data. As a caution, it can be easy to generate conflicting heuristics. While expert systems can help, their use represents an additional paradigm for users to learn and support. Other filters Aliasing is when a stray, high-frequen- cy signal confounds a sampled data signal. Sources include harmonics from rotating electrical motors, power transformers, and radio transmission. An anti-aliasing filter rejects the con - founding signal. In many industrial platforms, an anti-aliasing filter is a simple first-order filter with a time constant set slightly faster than the base scan rate for the I/O system. This will mostly eliminate the effects of any signals with periods faster than what the controller can respond to and prevent them from being "felt" as a slower, longer period disturbance. High-quality input cards with appro - priate anti-aliasing filters will elimi- nate radio effects for most common process signals. There are many filters in the loop that may attenuate noise from vari- ous sources. For example, a ther- mow ell acts as a filter to temper temperature fluctuations in the fluid when vapor and liquid are in trans- port. Lags in sensors, such as ion transport across the pH membrane, temper concentration fluctuations. Averaging in sample accumulation before analysis tempers fluctuation. The process engineer may have ad- justed a derivative filter, tuned a valve positioner, or added deadband on a controller output or actuator. The process engineer may have selected signal damping effects on an orifice dP transducer. Additionally, many sensors in industry today come with their own microprocessor providing selected features. Some include em - bedded data filtering for which some adjustment (i.e., tuning constant) is available to customers. While we acknowledge such diverse applications, this article focuses on techniques that are typically pro- grammed or configured into process control or data historian computers for which users have significant dis- cretion for their use and configuration. Perspectives l Filtering can be used to either temper noise or eliminate outliers. Use the right tool for the disparate applications. l If the signal is noiseless (and void of outliers), then there is no need to consider filtering. l Some applications (driven by regula- tory considerations) may also indi- cate no use of filtering. l If the noise level changes, then the user needs to adjust the filter factor to maintain the desired balance of noise attenuation to lag. l Statistical filters automatically adapt to changes in noise amplitude. l Filtering adds a lag or delay, which could impair control action, or re quire alternate controller tuning. l Diverse filtering methods can be used in combination, such as a me- dian filter to reject outliers, then the FoF to reduce noise. l Filter effects and options are on nearly every device. Recognize where these might be. n ABOUT THE AUTHORS Joseph S. Alford, PhD, is an automation consultant in the pharmaceutical in- dustry, previously completing a 35-year career at Eli Lilly in automating life sci- ence processes. He has authored or co-authored 45 publications, including book chapters, technical reports, and an ANSI/ISA standard. Alford is a member of InTech's advisory board, is an ISA and AIChE Fellow, and is a member of the Process Automation Hall of Fame. Brian M. Hrankowsky is a consultant en gineer at Eli Lilly and Company with 18 years of experience in industrial con- trols in the pharmaceutical industry. He has experience in controls applications in batch, continuous, and discreet manu- facturing systems on a variety of DCS, PLC, SCADA, and vision platforms. R. Russell Rhinehart is a professor emeri- tus at Oklahoma State University, with 13 years prior industrial experience. He is an ISA Fellow and member of the Pro cess Automation Hall of Fame. Rhine- hart is the author of three textbooks and six handbook chapters and maintains a Web site ( to provide open access to software and techniques. View the online version at RESOURCES Alford, J., et. al. (1999, April). "Real rewards from artificial intelligence." InTech. Alford, J., et. al. (1999, July). "Online expert-system applications; Use in Fermentation Plants." InTech. Rhinehart, R. R., Nonlinear Regression Modeling for Engineering Applica - tions: Modeling, Model Validation, and Enabling Design of Experiments, Wiley, 2016. Muthiah, N., and R. R. Rhinehart, "Evaluation of a Statistically Based Controller Override on a Pilot-Scale Flow Loop." ISA Transactions, Vol. 49, No. 2, pp 154-166, 2010. Although mathematical equations are the basis of conventional filters, human logic statements can be powerful additional ways to "clean up" data.

Articles in this issue

Links on this page

Archives of this issue

view archives of InTech - JUL-AUG 2018