Property Prediction and CAMD CHEN 4470 – Process Design Practice Dr. Mario Richard Eden Department of Chemical Engineering Auburn University Lecture No. 21 – Property Prediction and Computer Aided Molecular Design March 28, 2013

Property Prediction 1:2 Motivation Group Contribution Methods Experiments are time-consuming and expensive. How do we identify the components to investigate? Components of similar molecular structure have been found to have similar properties. Group Contribution Methods Predominant means of predicting physical properties for new components. Based on UNIFAC group descriptions Large amounts of experimental property data has been fitted to obtain the contributions of individual groups.

Property Prediction 2:2 Examples and Software

CAMD 1:3

CAMD 2:3

CAMD 3:3 Application Examples Water/phenol system: Toluene replacement Separation of Cyclohexane and Benzene Separation of Acetone and Chloroform Refrigerants for heat pump systems Heat transfer fluids for heat recovery and storage and many others

Aniline Case Study 1:7 Problem Description Conventional Approach During the production of a pharmaceutical, aniline is formed as a byproduct. Due to strict product specifications the aniline content of an aqueous solution has to be reduced from 28000 ppm to 2 ppm. Conventional Approach Single stage distillation. Reduces aniline content to 500 ppm. Energy usage: 4248.7 MJ No data is available for the subsequent downstream processing steps.

Aniline Case Study 2:7 Objective Reported Aniline Solvents Investigate the possibility of using liquid-liquid extraction as an alternative unit operation by identification of a feasible solvent Reported Aniline Solvents Water, Methanol, Ethanol, Ethyl Acetate, Acetone

Aniline Case Study 3:7 Performance of Solvent Liquid at ambient temperature Immiscible with water No azeotropes between solvent & aniline and/or water High selectivity with respect to aniline Minimal solvent loss to water phase Sufficient difference in boiling points for recovery Structural and EH&S Aspects No phenols, amines, amides or polyfunctional compounds. No compounds containing double/triple bonds. No compounds containing Si, F, Cl, Br, I or S

Also, higher and branched alkanes were identified as candidates Aniline Case Study 4:7 Results of Solvent Search No high boiling solvents found Also, higher and branched alkanes were identified as candidates

Aniline Case Study 5:7 Process Simulation

Aniline Case Study 6:7 Performance Targets and Results Countercurrent extraction and simple distillation. Terminal concentration of 2 ppm aniline in water phase. Highest possible purity during solvent regeneration

Aniline Case Study 7:7 Validation of Minimum Cost Solution

Oleic Acid Methyl Ester 1:3 Problem Description Fatty acid used in a variety of applications, e.g. textile treatment, rubbers, waxes, and biochemical research Reported solvents: Diethyl Ether, Chloroform Goal Identify alternative solvents with better safety and environmental properties. Volatile Flammable Carcinogen

Oleic Acid Methyl Ester 2:3 Solvent Specification Liquid at normal (ambient) operating conditions. Non-aromatic and non-acidic (stability of ester). Good solvent for Oleic acid methyl ester. Constraints Melting Point (Tm) < 280K Boiling Point (Tb) > 340K Acyclic compounds containing no Cl, Br, F, N or S Octanol/Water Partition coefficient (logP) < 2 15.95 (MPa)½ < δ < 17.95 (MPa)½

Oleic Acid Methyl Ester 3:3 Database Approach (2 Candidates) 2-Heptanone Diethyl Carbitol CAMD Approach (1351 Compounds Found) Maximum of two functional groups allowed, thus avoiding complex (and expensive) compounds. Formic acid 2,3-dimethyl-butyl ester 3-Ethoxy-2-methyl-butyraldehyde 2-Ethoxy-3-methyl-butyraldehyde Calculation time approximately 45 sec on standard PC.

Property Based Design Why Design Based on Properties? Many processes driven by properties NOT components Performance objectives often described by properties Often objectives can not be described by composition Product/molecular design is based on properties Insights hidden by not integrating properties directly Property Clusters Extension to existing composition based methods Reduces dimensionality of problem Enables visualization of problem Property estimation in molecular design via GC Unifying framework for simultaneous solution

Property Clusters 1:2 Property clusters are conserved surrogate properties described by property operators, which have linear mixing rules, even if the operators themselves are nonlinear. R.H.S.

Match clustering target Property Clusters 2:2 Feed Constraint Feasibility Region Analysis has shown that region boundary can be described by 6 unique points. Feasibility Necessary Condition Match clustering target Sufficient Condition Match AUP value of sink

Group Contribution Methods Group Contribution Methods (GCM) allow for prediction of physical properties from structural information 1st order, 2nd order, and 3rd order groups are utilized to increase the accuracy of the predicted properties

Molecular Clusters 1:5

Molecular Clusters 2:5

Molecular Clusters 3:5 2 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,9 C3 C2 C1 M1 G1 G2 G3 G4 Feasibility Region b1, the visualization arm, corresponds to the location of G1-G2 molecule

Molecular Clusters 4:5 1: CH3 Molecular Synthesis 2: CH2 3: CH3N 4: COOH CH3-(CH2)2-CH3N-COOH

Molecular Clusters 4:5 1: CH3 Molecular Synthesis 2: CH2 3: CH3N 4: COOH 5: CH3N-COOH CH3-(CH2)2-CH3N-COOH

Molecular Clusters 4:5 1: CH3 Molecular Synthesis 2: CH2 3: CH3N 4: COOH 5: CH3N-COOH 6: CH3-CH2 CH3-(CH2)2-CH3N-COOH

Molecular Clusters 4:5 1: CH3 Molecular Synthesis 2: CH2 3: CH3N 4: COOH 5: CH3N-COOH 6: CH3-CH2 7: CH3-(CH2)2 CH3-(CH2)2-CH3N-COOH

Molecular Clusters 4:5 1: CH3 Molecular Synthesis 2: CH2 3: CH3N 4: COOH 5: CH3N-COOH 6: CH3-CH2 7: CH3-(CH2)2 CH3-(CH2)2-CH3N-COOH

Formulation of Butyl methyl ether Molecular Clusters 5:5 The location of each molecular formulation is unique and independent of group addition path Formulation of Butyl methyl ether CH3-CH2-CH2-CH2-CH3O

Example: Molecular Synthesis Blanket Wash Solvent Design Solved as MINLP by Sinha and Achenie (2001) Problem Statement Design blanket wash solvent for phenolic resin printing ink Molecules designed from 7 possible groups, with a max. chain length of 7 groups Property Lower Bound Upper Bound Hv (kJ/mol) 20 60 Tb (K) 350 400 Tm (K) 150 250 VP (mmHg) 100 --- Rij 19.8

Blanket Wash Solvent 1:7 Visualization limits problem to three properties Heat of vaporization, boiling and melting temperatures are used, with vapor pressure and solubility used as final screening properties Property Prediction (GCM) Molecular Property Operators ,Y ref = 20 ,Y ref = 100 ,Y ref = 7

Blanket Wash Solvent 2:7

Blanket Wash Solvent 3:7

Blanket Wash Solvent 4:7

Blanket Wash Solvent 5:7 Feasible formulations from Visual Synthesis Application of feasibility conditions All formulations satisfy the first two necessary conditions M9-M11 fail to satisfy the AUP range of the sink

Blanket Wash Solvent 6:7 Feasible formulations from Visual Synthesis Application of feasibility conditions Checking property values with sink including Non-GC properties (VP, solubility), the sufficient conditions are satisfied for remaining formulations

Blanket Wash Solvent 7:7 Cyclical compound Ethers Ethers MEK Candidate molecules M1-M7 identified visually by the developed method correspond to solutions found by the MINLP approach used by Sinha and Achenie (2001) Although valid formulation, heptane (M8) is flammable hence not an ideal solvent Cyclical compound Ethers Ethers MEK

Integrated Design Approach Stream Properties & Unit Constraints Process Design Clusters Process/ Product Design Calculations Property Targets ClustersM Molecular Formulations Molecular Design Molecular Design

Example – Integrated Design Stream Characterization Objective To maximize the use of off-gas condensate and to minimize fresh solvent use to the degrease Stream Characterization Sulfur Content (S) Molar Volume (Vm) Vapor Pressure (VP)

Metal Degreasing 1:9 Degreaser Feed Constraints Property Lower Bound Upper Bound S (%) 0.00 1.00 Vm (cm3/mol) 90.09 487.80 VP (mmHg) 1596 3040 Property Operator Mixing Rules , S ref = 0.5 wt% , Vmref = 80 cm3/mol , VP ref = 760 mmHg

Visualization of Process Design Problem Metal Degreasing 2:9 Visualization of Process Design Problem VOC Condensation Data Sulfur content, density and vapor pressure data given for temperature range 480K-515K

Visualization of Process Design Problem Metal Degreasing 3:9 Visualization of Process Design Problem Conditions Condenser operates @ 500 K Feed Solvent must have zero sulfur content C 2 0.1 0.9 0.2 0.8 POINT A 0.3 0.7 Point A & B dictate property constraint targets 0.4 0.6 0.5 0.5 0.6 0.4 DEGREASER 0.7 0.3 0.8 0.2 POINT B 490 K 495 K 500 K 0.9 510 K 0.1 480 K 485 K 505 K CONDENSATE 515 K C C 3 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Metal Degreasing 4:9 Values from Process Design Visual Solution (%) Vm (cm3/mol) VP (mmHg) 102.09 1825.37 720.75 3878.66 Molecular Property Constraints Hv (kJ/mol) Vm (cm3/mol) VP (mmHg) 50 102.09 1825.37 100 720.75 3878.66

Property Prediction (GCM) Molecular Property Operators Metal Degreasing 5:9 Property Prediction (GCM) Molecular Property Operators 20 , = ref y 100 , = ref y 7 , = ref y Non-GC Property

Visualization of Molecular Design Problem Metal Degreasing 6:9 Visualization of Molecular Design Problem Molecular Fragments G1: CH3 G2: CH2 G3: CH2O G4: CH2N G5: CH3N G6: CH3CO G7: COOH

Visualization of Molecular Design Problem Metal Degreasing 7:9 Visualization of Molecular Design Problem Candidate Molecules M1 CH3-(CH2)5-CH3CO M2 CH3CO-(CH2)2-CH3CO M3 (CH3)3-(CH2)5-CH2N M4 CH3-(CH2)2-COOH M5 (CH3)2-CH3CO-CCL M6 -(CH2O)5- ring M7 CH3-(CH2)2-CH3N-COOH

Metal Degreasing 8:9 Formulations from Visual Design AUP Tb (K) Hv (kJ/mol) Vm (cm3/mol) VP (mmHg) M1 5.06 450.58 53.19 156.85 2078.98 M2 4.71 448.54 54.13 118.03 2163.90 M3 5.11 437.29 49.35 189.41 2692.07 M4 4.86 438.97 63.29 93.39 2606.12 M5 4.02 413.20 43.88 121.14 4241.48 M6 4.19 428.11 44.22 127.66 3208.12 M7 5.71 485.01 70.24 112.52 1037.99 Application of Feasibility Conditions All formulations satisfy first two necessary conditions M5 and M6 fail to satisfy sink AUP range M3 and M7 did not match Non-GC property value M1, M2 and M4 are valid solvent candidates

Visualization of Process Design Solution Maximization of Condensate Metal Degreasing 9:9 Solutions to Molecular Design Problem Visualization of Process Design Solution Maximization of Condensate 17.44 kg/min of condensate recycle is utilized 19.36 kg/min of 2,5-hexadione as fresh solvent

Summary Property Prediction and CAMD Can generate data for the simulation software in order to solve novel problems Allows for development of environmentally benign designs and components Systematic approaches, do not rely on rules of thumb. Utilizes process/product knowledge at selection level. Expands solution space for solvent design/selection Capable of identifying novel compounds not included in databases and/or literature Methodology has been proven through numerous application studies Powerful tool when used in an integrated framework

Other Business Next Lecture – April 4 Progress Report No. 3 Product engineering and Six Sigma SSLW pp. 662-678 Progress Report No. 3 Friday April 5 Remember to fill out the team evaluation forms