Neural network models were set up to take primary process inputs from PI and predict the values of paper strength properties in real time. Since the lab results weren’t reported until up to 20 minutes later, this allowed for immediate action by the operators. PI-Batch was used to sort production and model data into sellable paper sets and ActiveX controls were written to calculate averages and highlight cull limits for these units. The controls were displayed in ProcessBook with PI-API as the engine until server load proved to be a problem and the controls were rewritten using the PI-SDK.
Speaker
Fletcher Ashcraft
Fletcher has been contracted as PI Systems Analyst and Special Projects Programmer at Weyerhaeuser’s Pine Hill facility since 1996. He specializes in PI related custom programming utilizing VBA, MSAccess, MSExcel and VB 6 as containers and PI-API, PI-SDK and ODBC as data engines. He has authored several ActiveX controls and VB programs one of which, The Systems Optimization Program, was presented at the 2001 OSIsoft User’s Conference.