Download presentation
Presentation is loading. Please wait.
Published byRalf Phelps Modified over 8 years ago
1
1 CPAC: May 2005 A Roadmap for SAM A proposal to the Washington Technology Center
2
2 CPAC: May 2005 High-level vision of a full NeSSI system SAM’s role is to link data collected by NeSSI hardware to the standard process control system (dcs). SAM also generates QC information and translates the output of multivariate sensors into a form that can be handled by the dcs.
3
3 CPAC: May 2005 We are here NeSSI systems have the potential to overwhelm a digital control system The Sensor Activity Manager (SAM) has only been loosely defined If NeSSI is to get out of the starting blocks, SAM has to be able to deal with disparate sources and multivariate data to focus flexibly on specific applications
4
4 CPAC: May 2005 Responsibilities of SAM Manage registration and collect data from the suite of NeSSI sensors Perform consistency checks to validate the data Incorporate rules and control automating sample flow and recalibration Correct multivariate signals (spectroscopic and chromatographic) to account for drift Apply automated pattern recognition algorithms to interpret corrected sensor input Communicate full data and interpretation to a process experience database; and Send a reduce data feed to the DCS using the process standard OPC mechanism.
5
5 CPAC: May 2005 SAM must collect and store data Data is marshaled from the sensors into a data warehouse Organization is by time and date stamp Sensors all have an ID tag plus data in the form of a –Value –Vector –Array … SAM Sensors
6
6 CPAC: May 2005 SAM’s multivariate processing Because data can come from multivariate sources, SAM will need to reduce the communication to interpreted values (inferentials) and quality metrics. –Transformations –Classification –Regression Algorithms IPAK suite Macro driver DCS
7
7 CPAC: May 2005 3 Roles of Chemometrics in a Line Environment 1.Exploratory Functions Prior to building models When an unanticipated upset occurs Requires human intervention 2.Model Creation Requires human intervention 3.Routine Application of Models Needs to be configurable, application specific Fully automated
8
8 CPAC: May 2005 Sensor Application Manager SAM compiles data from all available sensors Part or all of this raw data can be sent to another computer on the network for –Archival –Exploratory work –Modeling Data can be processed to develop –Inferential measurements –Quality values –Diagnostics SAM DCS Sensors IPAK Commodity Data Specialty & QC Data Power User
9
9 CPAC: May 2005 Sensor Application Manager SAM assembles and transmits data from univariate sensors (commodity data) directly, functioning as a simple gateway SAM DCS Sensors IPAK Commodity Data Specialty & QC Data Power User
10
10 CPAC: May 2005 Sensor Application Manager SAM would process the entire sensor array to assess validity of individual sensors and determine overall parameters such as steady state SAM DCS Sensors IPAK Commodity Data Specialty & QC Data Power User
11
11 CPAC: May 2005 Sensor Application Manager Data from multivariate sources would be processed by IPAK to generate a specialty data feed to the DCS SAM DCS Sensors IPAK Commodity Data Specialty & QC Data Power User
12
12 CPAC: May 2005 Example SAM macro You get decision points, hierarchical models, plots, custom reports, full prediction control, no need to specify preprocessing Macro driver
13
13 CPAC: May 2005 The WTC project in summary LabView based Custom data warehouse ActiveX predictions InStep methods Customizable output Pirouette based InStep expert system Open to other tools Power User Macro driver Model development & troubleshooting Routine inferentials & reporting Data marshaling
Similar presentations
© 2024 SlidePlayer.com Inc.
All rights reserved.