Modeling & Simulation in Chemical Engineering

(CHE-F418)

Dr. Jay Pandey

BITS Pilani Assistant Professor, Chemical Engineering
Pilani Campus
 Module 1: Introduction to Modeling & Simulations

BITS Pilani
Pilani Campus
 Module 2: Classification of Mathematical Models

BITS Pilani
Pilani Campus
 Module 3: Development & Validation of a Model
L 1. Laws of Conservations/historical data
2. Numerical methods/Regression
3. Solvers
BITS Pilani 4. Optimization
Pilani Campus
5. Parameter estimation
 4
Module 4: Modeling of Chemical Engineering Systems
1. Reacting systems
2. Systems for mass transfer
3. Systems for heat transfer
BITS Pilani 4. Systems of fluid flow
Pilani Campus
 4
Module 5: Case Studies using computational methods
1. Process flow-sheet modeling (ASPEN)
2. Fundamental modeling (MATLAB)
3. Empirical/Data model (R-computing)
BITS Pilani
Pilani Campus
Pre-requisites for the Course

 Chemical Engineering Principles

 Numerical Methods in Chemical Engineering (Solvers &
Optimizers)

 Thermodynamics
 Reactor kinetics
 Transport phenomena
1. Heat transfer
2. Mass transfer
3. Fluid dynamics

BITS Pilani, Pilani Campus

Course Evaluative

BITS Pilani, Pilani Campus

Module 1: Introduction to Modeling &
Simulations

BITS Pilani, Pilani Campus

Evolution of Modeling

BITS Pilani, Pilani Campus

Breakthrough Phenomena (Phenomenological Model)

Newton’s Law of Motion, Einstein’s Theory of Sadi Carnot’s second law of TD, 1824
1687 Relativity, 1905

dS = dQ/T

Boyle’s experiment, 1646 Joule’s, 1846 Black Holes, 2008

PV= constant Q~W “Debatable”

dU = Q+W

BITS Pilani, Pilani Campus

Multiscale Modeling

 Thermodynamics

 Reaction Engineering

 Transport Phenomena

(Flow profiles: Velocity, Pressure,

Temp, Concentration)

BITS Pilani, Pilani Campus

Physics of Modeling

BITS Pilani, Pilani Campus

Materials Modeling_Applications

BITS Pilani, Pilani Campus

Multiscale Modeling_Applications

BITS Pilani, Pilani Campus

Importance of Modeling & Simulations

Ref.: Kunlanan at al. (2023), Design of PEM CO2 Electrolyzer for Electro-chemical Reduction of CO2 to Methanol

BITS Pilani, Pilani Campus

Mathematical Modeling?

“Mathematical Representation of any Physical/Real/Existing Systems”

Comprising of,

 Set of equations (algebraic or integral or differential)

 Set of variables (dependent/response & independent/predictor)

“DOF analysis (model evaluation)”

BITS Pilani, Pilani Campus

General form of a Mathematical System

BITS Pilani, Pilani Campus

Mathematical Modeling

DOF = (NV - NE)

For model simulation, NV= NE (well-specified)

NV>NE (under-specified)
NV < NE (over-specified)

*If NV>NE, the remaining NV – NE = NP variables (design variables) have to be

selected and fixed (as model parameters).

BITS Pilani, Pilani Campus

DOF analysis of a non-reacting system
Consider the perfectly mixed storage tank shown below. A liquid stream with volumetric rate Ff and density ρf flow
into the tank. The outlet stream has volumetric rate Fo and density ρo. Our objective is to develop a mathematical
model for the variations of the tank holdup, that is, the volume (V) of the fluid in the tank.
Carry out the DOF analysis for the system.

Overall mass balance

DOF = (NV - NE)

NV= 02 (L, Fo) & NE= 02
DOF = 0 (model specified)

BITS Pilani, Pilani Campus

DOF analysis of a reacting system

Overall mass balance (m3/s)

Component (B) mass balance (mol/s)

DOF = (NV - NE)

NV= 03 (L, Fo, CB) & NE= 03
Note: Reactant A is assumed to be in excess & not involved in the reaction.
DOF = 0 (model specified)

BITS Pilani, Pilani Campus

Module 2: Classification of Modeling &
Simulations

BITS Pilani, Pilani Campus

Classification_Mathematical Modeling

Rigid model:
(equilibrium…classical model
any variable at a fixed time is rig

Stochastic model

(quantum state…at a given time

BITS Pilani, Pilani Campus

Classification_Model System (Process)

Open-System (continuous process)

Closed-system (batch process)
Isolated-system

BITS Pilani, Pilani Campus

Classifications of Models: Based on Model Behavior

Steady Vs Unsteady Model

Stationary vs Transient Model

Static Vs Dynamic Model

Lumped (time, 0-D) Vs Distributed Model (time, 1-D, 2-D & 3-D)

Deterministic (rigid) Vs Stochastic (probabilistic) Model

BITS Pilani, Pilani Campus

Steady Vs Unsteady Model

BITS Pilani, Pilani Campus

Lumped Vs Distributed Model

“Set of ODE equations” “Set of PDE equations”

Lumped Model Distributed Model

BITS Pilani, Pilani Campus
Lumped Vs Distributed Model
 Lumped Model: No spatial (x, y, z) variation in temperature, pressure and
concentration. only variation is with time.
(e.g. reaction engineering problems, ODE in nature)

or
“The states (T, P, concentration) in lumped parameter systems are concentrated
in single point and are not spatially distributed”

 Distributed Model: Both time and spatial variation in flows/states (Temp,

pressure and conc.)
(e.g. fluid flow, mass and heat transport, PDE in nature)

BITS Pilani, Pilani Campus

Chemical Engineering_Fundamental Laws
Steady-state model Unsteady-state model

BITS Pilani, Pilani Campus

Module 3: Development & Validation of Model

BITS Pilani, Pilani Campus

Steps involved in Modeling & Simulations

Model Development
Model Solving
Model Analysis
Model Optimization
Model Validation

* Model: Fundamental/data-driven model

BITS Pilani, Pilani Campus
Approach to Modeling

Empirical modeling (theoretical modeling)…

Phenomenological model (observation cum experiment)

Fundamental (First-Principles) based modeling

Data-driven modeling (Predictive modeling)

BITS Pilani, Pilani Campus

Approach to Data Modeling
Input Output
 Collection of data (V1, V2, V3….Vn) (V1, V2, V3….Vn)
 Understanding data types (i.e. variables) Function
 Data reconciliation and identifying outliers
 Data processing
f (V1, V2, V3….Vn) = 0
 Model development, error analysis,
optimization V1= f(V2, V3….Vn)

V2 = f(V1, V3, ….Vn) ….so on

BITS Pilani, Pilani Campus

Approach to Data Analytics/Modeling

BITS Pilani, Pilani Campus

Variable-types

BITS Pilani, Pilani Campus

Machine Learning Algorithm

BITS Pilani, Pilani Campus

Data-Model Types

Linear Regression Model:

Y= βo+ β1*x1 + β2*x2 + β3*x3+….βn*xn

Non-linear Regression Model:

Y= βo + β1*x12 + β2*x1x2 + β32*x3+……β2n*xn

ANN model: “Black-box model”

BITS Pilani, Pilani Campus

Non-linear Regression Models

BITS Pilani, Pilani Campus

ANN Model Development

Activation Functions:
 Step function
 Linear function
 Rectified linear unit/leaky LU
 Sigmoid/logistic function
 Tangent hyperbolic function

BITS Pilani, Pilani Campus

Activation Function_ANN

BITS Pilani, Pilani Campus

Analysis of Data-Model (KPIs)
t-value:
Higher the t-value, the greater the
confidence & high reliability of the
predictive power of coefficient.

p-value:
Lower the p-value, the greater the
confidence & high reliability of the
predictive power of coefficient

CL = [1- (p-value)]

Standard deviation (σ) & standard

error (SE):

SE = (σ/n)

F-value: null vs alternate hypothesis

BITS Pilani, Pilani Campus

Regression

BITS Pilani, Pilani Campus

 Optimization of a Model
Step-1: Creating an Objective function (OF),

“OF is the Difference between model predicted values and actual experimental values (i.e. residual or
error)”
OF = sum (Model- Experiment)2

Step-2: Parameter estimation calculation, by taking derivative (i.e. slope) of the OF = 0,

Case study:

OF:

BITS Pilani, Pilani Campus

Approach to Fundamental Modeling

 Force/Momentum Balance
 Mass Balance  Simply, by applying laws of conservation to a model
 Energy Balance system (CV: open/closed)….right Model Assumptions
 Charge Balance
 Entropy Balance  Model type: Lumped or Distributed (0-D to 3-D)
 Lumped: ODE systems (steady/unsteady state)
 Distributed: PDE systems (unsteady/steady state)
BITS Pilani, Pilani Campus
Problem Statement: Force balance
A 2.5 m tall steel cylinder has a cross sectional area of 1.5 m2. At the bottom with a height of 0.5 m is liquid
water on top of which is a 1 m high layer of gasoline. This is shown in figure. The gasoline surface is exposed to
atmospheric air at 101 kPa. What is the highest pressure in the water?

BITS Pilani, Pilani Campus

Generalized Balance Equation

Note: Applicable to momentum-balance, mass-balance and energy-balance for any chemical

engineering systems

BITS Pilani, Pilani Campus

Module 4: Modeling of Chemical Engineering
Systems: Reaction Engineering

BITS Pilani, Pilani Campus

Mass/Mole Balance Equation

BITS Pilani, Pilani Campus

Isothermal CSTR Model with single reaction
Overall mass-balance

Component mass-balance

Model assumption: Homogeneous, Steady state reactor & Isothermal system.

Note: Since the system is well mixed, concentration CA and CB is equal to the
effluent concentration CAo and CBo.

BITS Pilani, Pilani Campus

CSTR Model with multiple reactions
Component wise mass-balance

Model assumption: Homogeneous, Steady state reactor & Isothermal system.

BITS Pilani, Pilani Campus

CSTR-in-Series Model
Component mass-balance for each tank

Model assumption: Homogeneous, Steady state reactor & Isothermal system.

BITS Pilani, Pilani Campus

Energy Balance Equation

BITS Pilani, Pilani Campus

Non-isothermal CSTR Model
Overall mass-balance

Component mass-balance

Energy-balance

BITS Pilani, Pilani Campus

Problem Statement

BITS Pilani, Pilani Campus

Heated Stirred Tank-in-Series Model
Energy-balance equation for each tank

BITS Pilani, Pilani Campus

Module 5: Solving System of Ordinary
Differential Equations (ODEs)

BITS Pilani, Pilani Campus

Classification of ODEs

Major emphasis is on first-order, linear, non-homogeneous type of ODEs (most common in

reaction engineering)

Where, y= state-variable, dependent (e.g., T, CA Pressure, Velocity)

x= independent-variable (e.g. time, volume, length)
P & Q are either constant or function of x

BITS Pilani, Pilani Campus

Problem Statements: Initial Value Problem

Note: Solve the ODE numerically and also using R (“desolve”) or MATLAB (“ode45”)

BITS Pilani, Pilani Campus

Solution of ODEs: IVP (initial value problem)
y(x=0)= yo , initial condition

P & Q are model parameters (constants) or function of x

ODEs are solved analytically & also using numerical

methods algorithms available in R or MATLAB (e.g.
desolve, ode-45, ode-23)

Classical examples of ODEs (mass & energy

balance eqn) in reaction engineering
(lumped-parameter models)

BITS Pilani, Pilani Campus

Problem Statements on IVP
(Applications in Chemical Engineering)

BITS Pilani, Pilani Campus

Problem Statement_IVP_Fluidized Packed Bed Catalytic
Reactor

BITS Pilani, Pilani Campus

Problem Statement_IVP_Biochemical Reaction

BITS Pilani, Pilani Campus

Problem Statement_IVP_Isothermal Plug Flow Reactor

PFR

PDE form for a PFR is converted to ODEs

BITS Pilani, Pilani Campus

Problem Statement_IVP_Fluid Flow

BITS Pilani, Pilani Campus

Problem Statement_IVP_Unsteady State Heat
Transfer

Solve the differential equation and find out the time required to reach the
steady-state conditions.

The steady state heat-conduction can be also represented in ODE form

and solved using IVP methods.
BITS Pilani, Pilani Campus
Problem Statement_IVP_Mass Transfer

Solve the first order differential equation of IVP and find out the graphically
show the variation of mole-fraction of liquid-water with r.
at r=0, yA(0)= 0.8,

Data given,
c = 5 mol/m3. DAB= 10-2 m2/s, NA= 20 mol /m2.s

Unsteady state diffusion can be also represented in ODE form

(specific) and solved using IVP.
BITS Pilani, Pilani Campus