New Vistas in Chemical Product & Process Design

Transcription

1 SPEED New Vistas in Chemical Product & Process Rafiqul Gani PSE for SPEED, Skyttemosen 6, DK3450 Allerod, Denmark (Head office: Bangkok, Thailand) 1

2 SPEED The big picture 2

3 SPEED Master of the planet earth how did we get there? Positive contributions to society Nutrition Fertilizers Electronics Communications, Entertainment Vitamins Fuel Mobility Healthcare Pharmaceuticals DetergentsColorants and coatings Plastics Clothing Housing Our survival depends on the products we make from the resources we have 3

4 SPEED The chemical product tree Question of what, why & when (how)? 4

5 SPEED Product Type Issues Examples Chemical product classification Single species Multiple species Small Large Formulated Functional Solvents, refrigerants APIs, surfactants, membranes Blended fuels, solvents, lubricants Detergents, personal care, healthcare, medicinal Devices Fuel-cells, microcapsules, hemodialysis device Key factors in design Basis of design Cost of production NA Process flowsheet (unit operations) Synthesis routes (chemistry) Cost and product functions Time to market (speed) Blend properties Microstructure and stability Product functions Device form and constituent materials Measure of performance Challenges in design Risks Closely related to process Translation of needs to molecular structure Closely related to application Translation of needs to properties Closely related to application Consumer satisfaction Delivering product functions defined by needs Consumer satisfaction Translation of needs to product material properties and configuration Large number of product Trial & error approach (non-optimal product) alternatives Feedstock availability Absence of engineering science knowledge Environmental impact and sustainability 5

6 SPEED The concepts 6

7 SPEED of single species products Find molecules with desired properties (also valid for mixtures) Pre-design "I want acyclic alcohols, ketones, aldehydes and ethers with solvent properties similar to Benzene" Interpretation to input/constraints A set of building blocks: CH3, CH2, CH, C, OH, CH3CO, CH2CO, CHO, CH3O, CH2O, CH-O + A set of numerical constraints (Start) A collection of group vectors like: 3 CH3, 1 CH2, 1 CH, 1 CH2O All group vectors satisfy constraints 2order group CH 3 CH 2 O CH 2 CH 3 CH CH 3 (Higher levels) CH 3 Refined property estimation Ability to estimate additional properties or use alternative methods Start of Postdesign CH 2 CH 2 CH 3 O CH 3 CH CH 3 CH 3 Group from other GCA method CH 3 CH 2 O CH CH 2 CH 3 Rescreening against constraints CH 3 CH 2 O CH CH 2 CH 3 Computer aided molecular design (generate & test) 7

8 Generation of Alternatives Groups as building blocks: CH 3 -, - CH 2 -, - OH, CH 3 CO -, (a set of about 180 groups available) Structural constraints (acyclic molecules) Size constraints Example: CH 3 CH 2 OH : Ethanol (3 groups) 8

9 Evaluation of Generated Molecules: Models Group contribution (GC) methods for property prediction GC MODELS Compound 1 Compound 2 Pure compounds Mixtures Examples: V 175 n C j j j Joback & Reid 1987 Examples: V V N Vc w M Vc z O Vc c c0 i 1i j 2 j k 3k i j k UNIFAC (VLE, LLE, SLE) COM ln ln ln RES i i i Gani, et al 2001, 2007, 2012, 2013,

10 SPEED Amino acids property modelling 10

11 Polymer : MINLP Applied to CAMD Fobj = [(P i P i* )/P i* ] 2 st 353 K < T g = [ (Y k n k )]/ [ (M k n k )] < 393 K 14 < = [ (M k n k )]/ [ (V k n k )] < 15 g/cm < W = [ (H k n k )]/ [ (M k n k )] < 00055) g H 2 O/g polymer) 2 y j 3 ; y j : 0 or 1 for j=1,7 2 y k n k 9 -CH2- ; -CO- ; -COO- ; -O- ; -CONH- ; -CHOH-; -CHCL- Note: objective function and constraints are non linear; n k, the optimization variables are integer (0-9) Polymer repeat units designed for bottle-stops; synthetic fibers; coatings, etc (CACE 2009) 11

12 Extension to process flowsheet synthesis Computer aided molecular design: CAMD CAMD versus CAFD Computer aided flowsheet design: CAFD 12

13 Sustainable Product-Process Development Raw material Applications: Biorefinery; CCU; CAMD; Waste-water treatment; 13

14 SPEED The problem definitions 14

15 SPEED Product-process & product-application relations Problems may have multiscale & multidiscipline features 15

16 SPEED Mathematical generic problem formulation Fobj = min {C T y + f(x, y, u, d, θ ) + S e + S i + S s + H c + H p } (1) 0 = h 1 (x, y) process constraints (Eq 2) 0 = P(f, x, y, d, u, θ) process model (Eq 3) θ = θ(f, x, y) product-property model (Eq 4) l 1 g 1 (x, u, d) u 1 process variable constraints (Eq 5) l 2 g 2 (x, y) u 2 molecular structure constraints (Eq 6) B x + C T y D process networks (Eq7) x: real-process variables; y integer-decision variables; u: process design variables; d: process input variables; θ: property; B, C, D coefficient matrices 16

17 SPEED Different problem formulations Eq 1: Objective function Eq 2: Process constraints Eq 3: Process models Eq 4: Property models Eq 5: Process variable constraints Eq 6: Molecular structural constrainst Eq 7: Processing networks Product Process (Eq1) Product Process Process simulation Process optimization Product evaluation Optimal product Product + implicit application Product-processapplication Product (Eqs4,6) Integrated Product-Process Proces s (Eqs2,3,5,7) Feed S 1 S N Simultaneous Product Process (multiscale & multidiscipline) 17

18 SPEED Concept of model based framework Methods and Tools Work Flow Information Flow Databases Knowledge base Model libraries Model generation Grand Product model Product (Eqs4,6) Product Process (Eq1) Product Integrated Product-Process Generate&Test (ProCAMD, SolventPro, CAPEC-Database) Optimization software (Solvers: CPLEX, BARON, ) Simultaneous or decomposed approach Process S 1 S N SimultaneousProduct Process (3) Process model selection Process (5) Objective function selection (Eqs2,3,5, Feed START Step 1: Problem Definition (1) Define the problem (2) Translate needs into target properties (3) Set the target property values Step 2: CAMD constraint selection (1) Structure model, groups & backbone (2) Property Thermodynamic model Step 3: CAMD formulation Step 4: Solution strategy (1) Selection of solution algorithm (2) Selection of MI(N)LP solver (3) Solution of MI(N)LP Solution found? Product needs Process simulation Process optimization Product evaluation Optimal product Product + implicit application Product-processapplication Target property values Economic/Sustainability/ Environment models Set of candidate groups Objective function Upper and lower bound of target properties Process/Product model equations MI(N)LP model List of candidates Report generation N END Y Zhang et al CACE,

19 Mathematical Problem Solution Target Where is Wally? Donde esta Waldo? Hvor er Holger? Our target optimal solution 1 9

20 SPEED Product (Eqs4,6) Product Process (Eq1) Integrated Product-Process S 1 S N SimultaneousProduct Process Process (Eqs2,3,5,7) Eqs 4, 6 with or without Eq 1 (product design and implicit performance verification) Feed Examples: Molecular design (single species products) Solvents Process fluids Surfactants Active ingredients Polymer repeat units 20

21 SPEED Product (Eqs4,6) Product Process (Eq1) Integrated Product-Process Process (Eqs2,3,5,7) S 1 S N Feed Examples: Molecular design (single molecular products) -Surfactants SimultaneousProduct Process Eqs 4, 6 with or without Eq 1 (product design and implicit performance verification) Homogeneous reactions 21

22 SPEED Surfactant - Product (1/3) Step 1: of a UV sunscreen, in the emulsified form, with a high sun protection factor The phenol ring is fixed as backbone structure in this case study Step 2: Convert needs-functions to properties Objective: Toxicity should be minimized, but higher than a lower bound Group selection: CH 3, CH 2, CH, C, ach, ac-oh, CH 2 COO, CH 3 O, CH 2 O, ac-o, OCH 2 CH 2 OH from 220 groups (Marrero and Gani, (2001)) Backbone selection: -C 6 H 4 OH Structural and property constraints: LC 50 : Fathead Minnow 96-hr exposure; Sp: Hildebrand solubility parameter; clp: cloud point; Fp: Flash point; Vm: Liquid molar volume at 298K 22

23 SPEED Surfactant (2/3) Step 3: Formulate the mathematical problem #Equations: 1,324,764; #Variables: 1,286,278 (1,286,266 Integer variables) Step 4: This design problem can also be solved directly using the GAMS CPLEX solver 23

24 SPEED Surfactant (3/3) Backbone 24

25 SPEED Product (Eqs4,6) Product Process (Eq1) Product Integrated Product-Process Process S 1 S N SimultaneousProduct Process Process (Eqs2,3,5,7) Eqs 2, 3, 4, 5, 6 with or without Eq 1 (product-process development; solvent designverification) Feed Examples: Molecular and/or mixture design with process application Solvents for separations and/or product recovery Process fluids for specific operations Reactive agents 25

26 SPEED Solvent based reaction-separation processes Generation of alternatives for the production of L-2-aminobutyric acid Objective of the study: Identify possible process hotspots to improve the process in terms of: Conversion to continuous manufacturing Solvent selection to improve productivity Reaction improvement Separation improvement 26

27 SPEED Solvent based reaction-separation processes Generation of alternatives for the production of L-2-aminobutyric acid Reaction type Database A Single phase system B Biphasic system Experimental data available Kinetic model is not available Solvent selection: Organic solvent that is, Immiscible with the aqueous phase Higher by-product selectivity Inhibits the enzyme 27

28 SPEED Solvent based reaction-separation processes Generation of alternatives for the production of L-2-aminobutyric acid Reaction type Database: Model development method: Multiphase reaction modeling (Anantpinijwatna et al 2016) Physical Equilibrium d A O 0 X AAD: i K18% i X A i R V Reaction and Mass Transfer R k C A BeA C A 2OA C C A A BeD L2 A Eq K Mass Balance N N dt 0 i N i O i N i ij j j N A i Experimental data 28

29 SPEED Solvent based reaction-separation processes Generation of alternatives for the production of L-2-aminobutyric acid Extractive solvent selection: Batch to continuous : Batch to continuous database Enzyme membrane reactor Facilitates enzyme stability and avoid the enzyme contact with the organic solvent in a continuous operation 29

30 SPEED Solvent based reaction-separation processes Generation of alternatives for the production of L-2-aminobutyric acid Objectives: Recover the main product Recover and recycle the extraction solvent (Hexane) and the reaction solvent (water) Recycle unreacted Benzamine Rank Process alternatives [SFILES] 1 (rabcdef)(spabcd/ef)(lme/f)[[(oe)]1(pf)](crd/abc)[(od)](dsab/c)[(oab)] (oc) 2 (rabcdef)(spabcd/ef)(lme/f)[[(oe)]1(pf)](crcd/ab)[(oab)](scsc/d)[(od)](scs/c)[2(os) ](oc) 3 (rabcdef)(spabcd/ef)(lme/f)[[(oe)]1(pf)](scd/sabc)[(od)](dss/abc)[(os)](oabc) F A B E F Phase separation 4 (rabcdef)(spabcd/ef)(lme/f)[[(oe)]1(pf)](dsab/cd)[(oab)](scsc/d)[(od)](dss/c)[2(os C A A: water )](oc) D B D A E B: BA C B F D A C: 2OA B D: AABA C E: BD C F: Hexane E Distillation 30

31 SPEED Solvent based reaction-separation processes Other biphasic reaction-separation systems 31

32 SPEED Product (Eqs4,6) Product Process (Eq1) Product Integrated Product-Process Desig n Process S 1 S N SimultaneousProduct Process Process (Eqs2,3,5,7) Feed Eqs 1-7: simultaneous product design & process application Examples: Molecular design (single molecular products) with process constraints Solvents for separations and/or product recovery Process fluids for specific operations Reactive agents 32

33 SPEED Simultaneous product and application process design Refrigerant and cycle design-1 Refrigeration cycle and fluid affect each other Therefore, product and process application considered Objective is to determine novel working fluids and mixtures that comply with current and upcoming regulations Requirements for substitution of existing fluids: New fluid should have same or improved cycle efficiency New fluid should have similar operational behavior New fluid should improve sustainability Relate process with product properties Find optimal product-cycle design Fluid Cycle Sustainable 33

34 SPEED Simultaneous product and application process design R134a is a commonly used refrigerant that is being phased out Product properties that affect the process cycle: Critical properties; Heat capacity; Compressibility; Accentric factor; Vapor pressure; Heat of vaporization These product properties influence the process cycle and its efficiency Also, other properties (enthalpy, entropy, fugacity) are needed for process cycle design Equation of state and for estimation of PVT relationship Refrigerant-cycle design: Step 1 Evaporator P Compressor Expansion Vavle Condenser 34

35 SPEED Simultaneous product and application process design Look for novel fluids that can contain C, H, F and Cl (up to 10 groups) 30 molecular groups are selected from M-G method Target properties are chosen for product and process: Target Property Lower bound Upper bound Molecular mass 110 g/mol Normal boiling point 250 K Critical temperature 350 K 400 K Critical pressure 30 bar 50 bar Thermal conductivity (liquid, Tb) 008 W/mK Step 2 Enthalpy of Vaporization at Tb 200 kj/kg ODP 0 GWP 1400 Atmospheric Lifetime * 14 years Number of groups 1 5 Number of functional groups Process needs 1 Lower bound 3 Upper bound Evaporation temperature K Condensation temperature K

36 SPEED Simultaneous product and application process design Molecular constraints n 2 v 2 i i i 1 n i 10 i 0 n i 3 Property constraints n i Mw i 110g / mol i i T ln b0 n T i b 250K i i 350K Tc0 ln nt i c 400K i i Refrigerant-cycle design: Step-3 (complete model-1) 1 30bar 2 Pc01 50bar n i P ci P c02 i T 2/ T c 01W / mk 1/2 2/3 T i i i b T c n Mw H T v 0 T 1 1 T c 0 1 T T c Process constraints kJ / kg f (Z ) 0 Z 3 Z 2 Z ( B 2 A B) AB V 1 V 1 V 1 V 1 f (Z ) 0 Z 3 Z 2 Z ( B 2 A B) AB V 2 V 2 V 2 V L3 L3 L3 L3 f (Z ) 0 Z Z Z ( B A B) AB Needed for efficient solution of cubic EoS [4]: f (Z V1 ) 0 f '(Z V 1 ) 0 f ''(Z ) 0 V 1 V 2 f (Z ) 0 f '(Z V 2) 0 f ''(Z V 2 ) 0 f (Z ) 0 L3 f '(Z ) 0 L3 f ''(Z ) 0 4] Kamath, RS, Biegler, LT, Grossmann, IE, 2010, Comput Chem Eng 34, L3 36

37 SPEED Simultaneous product and application process design Refrigerant-cycle design: Step-3 (complete model-2) Process constraints ln V1 ln L1 Compressor ln V 2 ln L2 H L3 (T 3, P 3 ) H L4 (T 4, P 4 ) 0 S V 2 (T 2, P 2 ) S V1 (T 1, P 1 ) 0 P 1, T 1 P 2, T2 P 1 P 4 P 2 P 3 T 1 T 4 P 1 P 0 P 2 08P c T 2 08T c 26565K T K 31315K T K Objective function Evaporator P 4, T 4 Expansion Vavle P 3, T 3 H Z,T, P H Z,T, P 1 V L3 3 3 max COP H 2 Z V 2,T 2, P2 H 1 Z V 1,T 1, P1 Condenser 37

38 roblem II Problem I SPEED Case study: Product-process design Cycle design for each molecules is given by Problem (II) r134a r152a 3,3,3- trifluoro propyne 3,3-diFluoro butene Results 2,2-diFluoro butane Mw 102 g/mol 66 g/mol 94 g/mol 92 g/mol 941 g/mol Tb 236 K 230 K 242 K 246 K 252 K Tc 374 K 366 K 369 K 392 K 389 K Pc 3732 bar 4675 bar 3813 bar 3927 bar 378 bar ω k 0082 W/mk 0097 W/mk 0086 W/mK 0093 W/mK 0092 W/mK ΔH v at T b J/mol J/mol J/mol J/mol J/mol ODP GWP (ASHRAE, 2005) (ASHRAE, 2005) Atmospheric lifetime 134 years (ASHRAE, 2005) 15 years (ASHRAE, 2005) - - COP

39 SPEED Product Process (Eq1) Product Process Examples: Formulated and/or blend design Liquid formulated products Product (Eqs4,6) Integrated Product-Process Process (Eqs2,3,5,7) S 1 S N Feed Tailor-made blends SimultaneousProduct Process Eqs 1, 4 & 6: mixture-blend design Emulsified products 39

40 SPEED Tailor-made fuel blends - 1 The main challenge involves how to identify the blends that satisfy the blend target properties with various types of additives Problems solved: Tailor made design of gasoline blends, diesel blends, jet-fuel blends and lubricant blends Example: Blending gasoline with different additives: impact of fuel consumption on the environment performances can be retained and improved Blend templates help to efficiently narrow down the search space of feasible chemicals to be added and obtain improved tailor-made blends in a faster, reliable and robut manner * Collaboration with Texas A&M in Qatar * To be presented at Session I-2, paper OL5, 16:40 by Sawitree Kalakul 40

41 SPEED Tailor-made fuels - 2 Task 1 Problem Defintion 11 Identify product needs 12 Translate needs into physico-chemical properties 13 Set target values Database Model Oil (fixed) Additives (22 selected) tailor-made blend 41

42 Nonlinear Program Linear Program SPEED Tailor-made fuels Task 2 Mixture/blend design 21 Pure component constraints (HHV, ρ, η, LC50, WtO2) 22 Miscibility constraint ( G mix ) 23 Linear constraints (HHV, ρ, η, LC50, WtO2) 24 Non-linear constraints (RVP, Tf) Miscibility test: ΔG Mix < 0 min or max s t ζ k LB NC NC i f obj x x i ζ k k i ζ UB x i 1 = 0 i=1 0 < x i < 1 (5 Eqs One for each prop-erty) min or max s t f obj x NC xi γ i P sat i (31095 K) = 1 RVP B i=1 NC x i γ i T f,i = T f,mix i=1 NC x i 1 = 0 i=1 x LB < x i < x UB 42

43 SPEED Tailor-made fuels Task 3 Rank blend candidates according to a selection criterion Task 4 Experimental verification Sample Density at 15⁰C (g/ cm ^3 ) Dynamic Viscosity(mPas) Vapor Pressure at 378⁰C (kpa) Target Values Experimental Model Experimental Model Experimental Model Main Ingredient Blend Blend Blend Blend Blend

44 SPEED What next? Product Process (Eq1) Product Process Some interesting developments: Product (Eqs4,6) Integrated Product-Process Process (Eqs2,3,5,7) S 1 S N Feed The chemical product simulator SimultaneousProduct Process Product-process development 44

45 SPEED Chemical product design simulator There is a need for a product simulator with similar and more useful features than a typical process simulator Based on available data, models, methods and analysis tools, the first chemical product simulator has been developed: ProCAPD Kalakul et al, VPPD-Lab: The Chemical Product Simulator, in Methods and tools for chemical product design, Computer Aided Chemical Engineering, 39, (chapter 3), 2017 Product design Use design templates (molecule products, formulated products, blended products, emulsified products and devices) Product analysis Predict and analyze product behavior (identify important product properties) Product performance Simulated product performance through virtual application experiments Product search Search for data, models, properties, products, devices, etc 45

46 SPEED Product (Eqs4,6) Product Process (Eq1) Product Integrated Product-Process Process S 1 S N Simultaneous Product Process Process (Eqs2,3,5,7) Eqs 1-7: simultaneous product design & process application decompositionbased solution approach Feed Examples: Molecular design (single molecular products) with process constraints Solvents for separations and/or product recovery Hybrid process design with membranes Process fluids for specific operations Reactive agents 46

47 SPEED of hybrid modules Which is the product design problem? x D1 Permeate (X) x D1 x F CD 1 x F VP x D2 6 x B1 Feed (I) PV x B2 Retentate x D1 x F CD 2 x F Permeate Raffinate Multiscale5 modelling x B1 x F1 RD x D1 x F1 3 RD CD 4 x D2 x F2 x F2 x B1 x B1 x B2 Close to 50% or more energy reduction compared to original process achievable 47

48 SPEED Collaborative effort needed! REALITY VIRTUAL REALITY System under study Provide realistic model parameters Model system Experiments Guidance and insights for experiments Simulations Optimized design? Verify theoretical solutions Optimized design? Solution approaches Integrated multiscale modeling, experiments and synthesis Ability to find predictiveinnovative solutions 48

49 Problem from 60 Years Ago Non-scumming soap bar (world s bestselling soap bar) Launched by Lever Bros (US division of Unilever) in 1955 Attributes comparable to ordinary soap bars Firmness Lather Rate of wear Slipperiness Mildness No unpleasant odors No cracking from wet/dry cycles Michael Hill M Hill & Associates Mahwah, NJ Processable on ordinary soap bar line: model-based system yet to be developed

50 SPEED Textbook: Process-Product (4th edition) Published by Wiley Suitable for BSc, MSc level teaching of Product design Process design Integrated product- process design 50

51 SPEED The academic tree of Prof Rafiqul Gani 51