Live from WBF 2008-Modeling the Wild APC for Batch

Advanced Process Control for Chemical Batch Reactors Wim Van Brempt of IPCOS and David R. Vinson, from Air Products & Chemicals Advanced Process Control: MPC superior to PID control Variance reduction Multivariable Predictive Optimization Optimization = Business Case (he showed a slide I can't reproduce illustrating  how MPC(Model Predictive Control) works) APC in Continuous Processes: Commodity tool Black Box models  Step Response Models  Nonlinear extensions  e.g. for polymer grades Some observations about Batch Process Control: Batch Processing enables flexible production and delivers opportunity to compete with low cost countries … at the cost of more complex operations and control solutions Standard Advanced Process Control offers no technical solution for Batch processes Multi-product reactors have only limited production time/product /year requiring a short payback time of a control project Product Quality is only measured when the batch is completed, making within-the-batch quality control not very straightforward Challenges for a Batch Advanced Process Controller 2 Major technical challenges: Provide accurate process models in reasonable time Upgrade standard "linear" APC to enable more complex nonlinear models (he showed a slide that compares the batch vs continuous modeling approach) BATCH Modeler: a Hybrid Modeling Technology-- Heat Balance Modeling for Batch & Fed-Batch Reactors  Decrease modeling effort by preconfigured submodels  (segmented) Jacket, (segmented) Cooling Coil, Condenser,…  (non) boiling batch (autoclave/atmospheric)  First Principle modeling of heat exchange units  Parametric modeling of reaction heat based on logged data Parametric Part  Non linear surface map  Built-in knowledge  Largely based on historical data Result  Accurate model for control requiring limited modeling time  Easily extendable towards other reaction products  Short commissioning time  Physical based model allows interface between Chemical Engineer & Control Engineer. Chemical Engineer may become model owner enabling long term maintenance support. BATCH Modeling: Models trained based on historical data and limited tests on ±2 batches Fast and accurate modeling required, limiting engineering time to strict minimum! (enabling a short project payback time, needed for multi-product reactors) Models predict:   Temperature,   Heat of Reaction   Conversion   Unreacted feed material   Adiabatic Temperature Intra-Batch and Inter-Batch Solutions Intra-Batch actions "within the batch" and "during the batch" Inter-Batch actions "in between batches" and "in between subsequent batches" Observer Technology: Goal = on-line use of the BATCH Model Keep your model in track with reality – Online model adaptation  Technology integrated in online model  Follow up fouling, catalyst deactivation from batch to batch APC can optimize the intra-batch control solution. Can your batch run faster? APC can tell you. (he provided an application example screenshot) Rules: Intra-batch, run on constraints and stick to references Inter-batch, modify constraints and recipes to fulfill quality constraints Integration into existing assets Separate PC connected via OPC Embedded in BATCH Sequencing & LIMS Automatic startup when relevant phase has started Intelligent fallback strategy InterBatch Control integrated with LIMS Operator Screen for major functions Here's a typical APC Project Schedule: Presentation-->Benefit Analysis--> Kickoff-->Installation-->Pre-Test Logging-->Modeling-->Configuration-->FAT-->Commissioning-->SAT-->Warranty