Development of a Systematic Synthesis and Design Methodology to Achieve Process Intensification in (Bio)Chemical Processes Philip Lutze a, Rafiqul Gani b, John M. Woodley a a PROCESS, Center for Process Engineering and Technology b CAPEC, Computer Aided Process Engineering Center Department of Chemical and Biochemical Engineering, Danmarks Tekniske Universitet, DTU Soltofts Plads, Kgs. Lyngby, DK-2800 CAPEC
Message Integration of Process Intensification and Process Synthesis Tools can enable the quantitative Selection and Implementation of new PI operations 2
Outline of the Presentation Introduction Motivation? Methodology? Case Study Neu-5-Ac Conclusions & Future Work 3
Importance of Process Intensification? Chemical and bio-based industry already faces and will face enormous challenges to achieve and/or respond to: Establish sustainable production Raw Materials Product(s) Survive global competition Quicker changing markets Utilities Process Future processes need to be: - Economic - Energy efficient - Raw material efficient - Safe Waste - Flexible - Renewable Demands for innovative products One option: Use of Process Intensification Only a limited number of PI implemented: Harmsen, 2010 Reverse Flow Reactor, Reactive Distillation, Dividing Wall Columns, 4
Motivation - Development of intensified processes Case-Based: Select one PI unit Trial & Error Approaches: Experimental-based strategies Simulation-based strategies Use of specific sub-methodologies: For reactive-separation schemes (Schembecker et al., 2003) Integration of distillation columns (Errico et al., 2009) Reactive distillation (Huang et al., 2010) Where is PI needed? How to select the PI unit? Near optimal design? No benchmark against other PI options? Resource consumption? Where is PI needed? How to select the PI unit? No benchmark against other PI options? Use of a general systematic synthesis/ design methodology to achieve PI (Lutze et al., 2010) 5
Methodology - Concept Process Intensification Process Synthesis Creates a large number of possibilities Feasible option which mostly improve the process Improvement Feasible Identified potential options Search space Systematic synthesis methodology to achieve PI through Decomposition approach in hierarchical steps Process Synthesis is the systematic strategy to identify the optimal path to reach a product in desired quantity & quality with respect to defined constraints on the process. 6
Methodology Decomposition approach in 6 hierarchical steps... 1.1 Give Objective: Min/Max F obj =Σf j (Y, X, d, Θ) (1) s.t. Y, X, d, Θ and, 1.2 Define/ Translate Scenario/Specifications into Constraints: Logical constraints: g Log (Y) 0 (2) Structural constraints: g Str (Y) 0 (3) Process Model: h p (X, d, Θ)=0 (4) Operational constraints: g Op (Y, X, Θ) 0 (5) 1.3 Define additional Performance Measure for Screening: Performance criteria: P l (Y, X, Θ) P Target,l (Y, X, Θ) 0 (6) 7
Methodology Decomposition approach in 6 hierarchical steps... 2.1 Collect data of the base-case design/process 2.2 Analyze Base-Case Design to identify limitations/bottlenecks: Algorithms consisting of: Calculation and analysis of F obj and P l of the base case design Analysis of the flowsheet by mass & energy indicators 2.3 Analyze limitations/bottlenecks: 2.3.1 Analyze limitations/bottlenecks by property analysis: Pure component analysis Reaction analysis (if present) Mixture analysis 2.3.2 Identify key phenomena responsible for limitation/bottleneck: Rules to link knowledge from 2.2.1 and 2.1 8
Methodology Decomposition approach in 6 hierarchical steps... 2.3 Retrieve PI technologies from a PI Knowledge-Base: Keywords: Process System, Reaction System Key phenomena Identified Limitation/ Bottleneck in a task Objective 2.4 Pre-Screen for feasibility: Determine operational window for each PI technology Compare with properties of the system 9
Methodology Decomposition approach in 6 hierarchical steps... Objective: Derive simple model equations for calculation of P l Process Model: h p (X, d, Θ)=0 (4) 3.1 Retrieve generic models from a model library 3.2 If model not existent, derive model Use of systematic model development strategies (Hangos&Cameron, 2001, Heitzig et al., 2010) 3.3 Validate model(s) 3.3.1 If Validation sucessful, store model in model library 3.3.2 If Validation not sucessful, remove option from search space 10
Methodology Step 1-3 4.1 Derive superstructure Use of generic superstructures from PI knowledge base Use of synthesis rules taken into account input/ output data of PI technologies 4.2-4.3 Screen sequentially by logical & structural constraints Logical constraints: g Log (Y) 0 (2) Structural constraints: g Str (Y) 0 (3) Fix all binary Y s Set of flowsheet options 11
Methodology # 1 # 2 5.1 Screen for operational constraints Simple Process Model: h p (X, d, Θ)=0 (4) Operational constraints: g Op (Y, X, Θ) 0 (5) 5.2 Screen by performance criteria and/or F obj Calculate and Rank: Performance criteria: P l (Y, X, Θ) P Target,l (Y, X, Θ) 0 (6) and/ or F obj =Σf j (Y, X, d, Θ) (1) 12
Methodology # 1 # 2 6.1 (Only for high number of remaining options) Solve MINLP problem with simple models to determine most promising options 6.2 Optimize most promising options to determine the best PI option Solve complete process synthesis problem (Equations 1-5) 6.3 Validation by rigorous simulation 13
Methodology Decomposition approach in hierarchical steps... 2.5 10 11 9.7 10 6 159/59 24 3 1 Step 2.1-2.3 Identify potential PI options Step 2.4 Pre-Screen for matching operational windows Step 4.2 & 4.3 Logical and structural constraints (Eqs.2&3) Step 5.1 Operational constraints (Eqs.4-5) Step 5.2 Screening by performance criteria (Eqs. 6 and/or 1) Step 6 Reduced Optimization Problem (NLP) at unit operation level (Eqs. 1-6) 14
Methodology Associated Tools PI knowledge base tool: 121 PI technologies Example for application in step 2 2.3 Collect potential PI equipment through keywords such as Limitations in current design Process System Target/ Potential Improvement 2.4 Pre-Screen for feasibility via Necessary conditions for application of each PI equipment 15
Example Two-Step Reaction Process data of existing process Step 1: Define problem Production of Neu5Ac Important pharmaceutical intermediate Building Block for Oligosaccharide-Production Production from Glucosamine and Pyruvate in 2 reactions Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3: Select & develop models Step 4: Generate feasible flow-sheet options Reaction 1: Epimerization OH OH CH2OH O AcHN A OH epimerase HOH2C OH OH B NHAc O OH A: GlcNAc: Glucosamine; B: ManNAc: Manosamine; Step 5: Fast screen for process constraints based on shortcut models Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option Reaction 2: Aldolase reaction HOH2C NHAc O OH OH H3C OH O C COOH aldolase HO OH OH AcHN o OH OH COOH C: Pyr: Pyruvate; D:NeuAc: Neuraminic acid; B C D 16
Example Base-Case design: Process Process data of existing process Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3: Select & develop models Production of Neu5Ac Base case design by Mahmoudian and coworkers: 5 Step process Reaction 1: Alkaline catalyzed by NaOH (Alk1) Reaction 2: Enzymatic catalyzed by Neu5Ac Aldolase (E2) Purity > 99% Productivity η 0 = 0.25 g l -1 d -1 Step 4: Generate feasible flow-sheet options Step 5: Fast screen for process constraints based on shortcut models Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option Mahmoudian et al., Enzyme Microb Technol 20, 393-400 (1997) 17
Example Production of Neu5Ac Step 1.1-1.4: Define Problem Objective: Maximize productivity η Process data of existing process Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3: Select & develop models Step 4: Generate feasible flow-sheet options Step 5: Fast screen for process constraints based on shortcut models Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option max F Obj = η/ η 0 = f (Y, X, d, Ф) with m 1 D V0 t Constraints such as: and η 0 = 0.25 g l -1 d -1 Amount and quality of product D, Usage of less than 4 process steps, Usage of mature PI unit operations. Additional Performance Metric for Screening in Step 4: Simplification, Energy, Waste Additional Performance Metric for Screening in Step 5: Time-Yield in the reaction steps: Overall Product Yield: R 0.75g g d 1 1 18
Example Production of Neu5Ac Step 2.1: Identify Bottlenecks/Limitations of the Base-Case Calculation of performance criteria: Process data of existing process Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3: Select & develop models Step 4: Generate feasible flow-sheet options Step 5: Fast screen for process constraints based on shortcut models Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option 1. Low Efficiency in both reactions ε Reaction1 = 0.2 ε Reaction2 = 0.9 2. Large Solvent Usage / Waste generation 5:1 (Solvent : Volume), Excess of reactant C 3. Energy demanding separations (purification) Evaporation of 70 L Water / kg Product 19
Example Process data of existing process Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3: Select & develop models Step 4: Generate feasible flow-sheet options Step 5: Fast screen for process constraints based on shortcut models Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option Production of Neu5Ac Step 2.2: Analyze Bottlenecks/Limitations of the Base-Case Characteristics of the Base-Case Design: 1. Low Efficiency in both reactions 2. Large Solvent Usage / Waste generation 3. Energy demanding separations (purification) Reaction Analysis Pure component Analysis Mixture Property Analysis 1. Unfavorable equilibrium 1. Substrate and Product Inhibition 2. Excess of Substrates 2. Substrate C and Product D with similar pka-value 3. Diluted System Reaction Kinetics: Zimmermann et al., 2007 20
Example Process data of existing process Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3: Select & develop models Step 4: Generate feasible flow-sheet options Step 5: Fast screen for process constraints based on shortcut models Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option Production of Neu5Ac Step 2.3: Identify PI Equipment Limiting Phenomena: Both reactions Keyword search: Limitations, Process system, Reaction, PI KBase Unit operations of base case PI technologies following two principles: PI by integration of both reactions PI by integration of reaction(s) and separation PI technology: (1) One-pot-reactor (18) Reactors integrated with a LL Extraction, Distillation, Evaporation, Membrane, Pervaporation, Adsorption, Crystallization, Stripping, Chromatography, Absorption, Precipitation, Comminution, Condensation, Distillation- Pervaporation, Distillation-Membrane, Spinning-disc reactor/ Rotating Packed Bed Reactor, Rotating Annular Chromatographic reactor, 21
Example Production of Neu5Ac Step 2.4: Pre-Screen for Feasibility Process data of existing process Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3: Select & develop models Step 4: Generate feasible flow-sheet options Step 5: Fast screen for process constraints based on shortcut models Potential Max. Number of process flowsheet options: 1.2 10 12 PI KBase Screen for necessary conditions, operational windows and maturity Example(s): Reactive Condensation: Phases: Vapor Liquid, Operational window: T,p: Between highest boiling point and lowest boiling point in the system Reaction: in Gas phase Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option One-Pot-Reactor: Phases: Liquid, Liquid-Liquid, Vapor-Liquid,. Operational window of reactions: T, p: moderate temperature at 1 bar ph-value: critical only for alkaline and enzyme 22
Example Production of Neu5Ac Step 2.4: Pre-Screen for Feasibility Process data of existing process Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3: Select & develop models Step 4: Generate feasible flow-sheet options Step 5: Fast screen for process constraints based on shortcut models Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option Potential Max. Number of process flowsheet options: 1.2 10 12 PI KBase Screen for necessary conditions, operational windows and maturity Non-intensified technology for external coupling: (Number of equipment): (7) Precipitation, Evaporation, crystallization, chromatography, Spray-Dryer, Centrifugation, LL-Extraction (5) Catalytic Reactors (Different catalysts: Reaction 1: alkaline: alk1,alk2; enzymatic: E1, Reaction 2: Enzymatic: E2, E22) PI technology: (5) One-pot-reactor (E1/ E2, E11/ E22, alk1/e2, alk1/ E22, WC), (10) All reactors integrated with a LL extraction, 27 Equipments Max Number of process flowsheet options: 9.7 10 6 23
Example Production of Neu5Ac Step 3: Select and/or Develop models Step 3.1. Retrieve models from a model library: Process data of existing process Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3.2. Derive models A superstructure of the process containing generic models based on mass& energy balances has been developed Step 3: Select & develop models Step 4: Generate feasible flow-sheet options Step 5: Fast screen for process constraints based on shortcut models Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option 24
Example Production of Neu5Ac Step 3: Select and/or Develop models Step 3.2. Derive models Process data of existing process A superstructure of the process containing generic models based on mass& energy balances has been developed Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3: Select & develop models Step 4: Generate feasible flow-sheet options and Step 5: Fast screen for process constraints based on shortcut models Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option 25
Example Production of Neu5Ac Step 3: Select and/or Develop models Example: Process data of existing process Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3: Select & develop models Step 4: Generate feasible flow-sheet options Step 5: Fast screen for process constraints based on shortcut models Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option 26
Number of process options in search space Example Production of Neu5Ac Step 4: Generate feasible flowsheet options Max number of process options: Process data of existing process Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3: Select & develop models Step 4: Generate feasible flow-sheet options Step 5: Fast screen for process constraints based on shortcut models Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option Step 4.1 Screening through logical constraints: The following integration schemes are considered: R-S, R-S-S, R-R-S, R-S-R-S, R-R-S-S ; Product D is formed only one possible configuration for each unit operation Step 4.2 Screening through structural constraints: Simplification, 9.7 10 6 1900 e.g. Do not connect two One-pot-Reactors 452 Efficiency, e.g. Deactivation of enzyme (R1(alk 2) -> R2(E2/E22)) 83 Energy, e.g. Do not evaporate water, add water, evaporate again 63 Waste, e.g. Do not use two different solvents 59 27
Example Production of Neu5Ac Step 5.1: Fast Screen for Process Constraints 1. Most promising with respect to reaction time yield: Reaction simulation for: Same initial concentrations & enzyme amount Process data of existing process Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3: Select & develop models Step 4: Generate feasible flow-sheet options Step 5: Fast screen for process constraints based on shortcut models Reaction configuration Reaction productivity [g g -1 day -1 ] OPRE (E1/E2) 0.087 OPRE (E11/E22) 0.865 OPRE (WC) 0.793 OPR (E1/E2) 0.115 OPR (E11/E22) 1.074 OPR (WC) 0.985 OPR(alk1/E2) 0.107 R1(alk1) - Chryst Enrich R2(E2) 0.050 R1(alk1) - Chryst Enrich R2(E22) 0.072 R1(alk2) - Chryst Enrich R2(E2) 0.262 24 process options remaining Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option R1(alk2) - Chryst Enrich R2(E22) 0.397 R1(E1) - Chryst Enrich R2(E2) 0.060 R1(E1) - Chryst Enrich R2(E22) 0.072 R1(alk1) - R2(E2) 0.031 R1(alk1) - R2(E22) 0.048 R1(E1) - R2(E2) 0.031 R1(E1) - R2(E22) 0.048 One-Pot Reactive Extraction; One-Pot-Reactor; Crystallization; Reactor; 28
Example Production of Neu5Ac Step 5.2: Benchmark with Performance Metric and F obj 1. Most promising with respect to reaction time yield: 24 2. Calculate performance metric yield and select most promising: 11 3. Calculate F obj and select most promising: 3 Process data of existing process Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Step 3: Select & develop models Step 4: Generate feasible flow-sheet options Step 5: Fast screen for process constraints based on shortcut models Step 6: Minimize objective One-Pot Reactive function & validate most promising by experiments Extraction; One-Pot-Reactor; Optimal intensified, feasible process option LL-Extractor; Chromatography; Crystallization; Reactor; Precipitation Process description Yield [%] Fobj [-] #1 OPRE (E11/E22) - Cryst Purif 45.9 161.7 #3 OPR (E11/E22) - Cryst Purif 60.8 214.0 #5 OPRE (E11/E22) LL - Cryst Purif 41.3 140.6 #9 OPR (E11/E22) LL - Cryst Purif 57.0 178.0 #10 OPR (E11/E22) LL - Chrom Purif 42.8 168.0 #13 OPRE (E11/E22) Evap - Cryst Purif 52.2 177.5 #15 OPR (E11/E22) Evap-- Cryst Purif 64.8 220.4 #17 OPR (E11/E22) Prec - Cryst Purif 64.8 227.4 #19 OPR (E11/E22) LL Evap - Chrom Purif 47.5 163.3 #21 OPR (E11/E22) LL LL Cryst Purif 51.3 155.3 #22 OPR (E11/E22) LL LL Chrom Purif 42.8 147.0 29
Example Production of Neu5Ac Step 6: Solve reduced optimization of most promising options Process data of existing process Step 1: Define problem Step 2: Collect data & identify bottlenecks/limitations to collect feasible PI equipment/ strategies for each task Objective: Maximize productivity η max F Obj = η/ η 0 = f (Y, X, d, Ф) m 1 with D and V t 0 Step 3: Select & develop models Step 4: Generate feasible flow-sheet options Step 5: Fast screen for process constraints based on shortcut models Step 6: Minimize objective function & validate most promising by experiments Optimal intensified, feasible process option Optimization Variables: Inlet concentration of reactants A and C Process Option #3 Process Option #15 Process Option #17 na,0 [mol] / l 0.932 0.932 1.3 nc,0 [mol] / l 2.185 2.185 2.05 Fobj [-] 324.8 314.0 334.5 30
Conclusions PI has the potential to improve processes. Development of a systematic synthesis methodology to achieve PI. It has the following main contributions: Efficient handling of large number of options through decomposition approach in hierarchical steps Quantitative reasoning/ screening of options Knowledge-Base tool Application of the developed unit-operation based framework to case studies shows good results Production of H2O2 Production of HMF Production of Neu5Ac Production of Cyclohexanol 31
Current & Future Work Case-Based: Select one PI unit Trial & Error Approaches: Experimental-based strategies Simulation-based strategies Use of specific sub-methodologies: For reactive-separation schemes (Schembecker et al., 2003) Integration of distillation columns (Errico et al., 2009) Reactive distillation (Huang et al., 2010) Where is PI needed? How to select the PI unit? Near optimal design? No benchmark against other PI options? Resource consumption? Where is PI needed? How to select the PI unit? No benchmark against other PI options? Use of a general systematic synthesis/ design methodology to achieve PI (Lutze et al., 2010) How to achieve PI beyond currently existing PI units? Use of a phenomena-based synthesis/ design methodology 32
Current & Future Work Methodology for phenomena-based synthesis/design Similiar strategy as in the unit-operation based method Managing the complexity by application of the decomposition approach Exploit similiarity to Computer-Aided Molecular Design Molecules Similiarity Processes Groups Unit operations Atoms Phenomena C H O 33
Conclusions PI has the potential to improve processes. Development of a systematic synthesis methodology to achieve PI. It has the following main contributions: Efficient handling of large number of options through decomposition approach in hierarchical steps Quantitative reasoning/ screening of options Knowledge-Base tool Application of the developed unit-operation based framework to case studies shows good results Production of H2O2 Production of HMF Production of Neu5Ac Production of Cyclohexanol Development of a phenomena-based synthesis/design methodology tackles limitation of pre-defined unit operations and opens up to achieve even higher benefits by using PI 34
Thanks a lot for your attention Contact: pil@kt.dtu.dk CAPEC