Process Modelling, Simulation & Computer-Aided Process Engineering (PMS & CAPE)

A comprehensive computational repository dedicated to Chemical Process Modelling & Simulation (PMS) and Computer-Aided Process Engineering (CAPE). This repository houses mathematical modeling from first principles, iterative numerical algorithms, machine learning / deep learning for thermophysical property prediction, pinch analysis for energy integration, shortcut and rigorous distillation column design, and industrial heat exchanger rating using Aspen HYSYS & EDR.

Portfolio Overview & CV Highlights

This repository contains academic and industrial-grade engineering projects completed as part of advanced chemical engineering coursework (CH620003: Process Modelling and Simulation and CAPE Laboratory at IIT Kharagpur).

Core Competencies Demonstrated:

First-Principles Numerical Simulation: Rigorous stage-by-stage MESH equation modeling, tridiagonal matrix solver (Thomas Algorithm), Newton-Raphson non-linear root finding, Simultaneous Correction (SC) methods.
Process Data Science & Scientific ML: Closed-form Normal Equation regression, 3D loss surface optimization, deep neural network (MLP) architecture design in TensorFlow/Keras for thermodynamic property prediction with sparse training sets (10% data).
Process Energy Integration & Optimization: Pinch Technology, composite curves ($T\text{-}H$ diagrams), Problem Table Algorithm (temperature cascade), minimum utility targeting, and grid diagram HEN synthesis saving 1.8 MW of energy.
Commercial Process Simulators: Industrial thermal design, rating, fouling analysis, and re-rating/retrofitting of shell-and-tube heat exchangers in Aspen HYSYS and Aspen EDR.

Important Project Reports (Ready for CV/Portfolio)

Below is the structured breakdown of the primary engineering projects in this repository, including direct links to their detailed reports, mathematical models, code files, and key quantifiable outcomes.

1. Multicomponent Multistage Distillation Simulation (Wang-Henke BP Method)

Domain: Rigorous Separation Processes & Numerical Modelling
Primary Report: 22CH30008_PMS_A1_Report.pdf (Section 4)
Hand Calculations & Verification: C_a hand iteration (1).pdf
Source Code: C_b.py, C_c.py, test.py

Problem Summary:

Simulation of a 3-stage (extendable to 10 stages) distillation column with a total condenser and partial reboiler at 1 atm separating a ternary alcohol feed: 60 mol% Methanol, 20 mol% Ethanol, 20 mol% n-Propanol ($F = 1000\text{ kmol/h}$, $D = 600\text{ kmol/h}$, $L_0 = 2000\text{ kmol/h}$).

Key Technical Contributions:

Formulated full MESH equations (Material, Equilibrium, Summation, Heat balances) for multicomponent vapor-liquid equilibrium (VLE).
Implemented an object-oriented Distillation solver in Python utilizing the Thomas Algorithm (TDMA) to invert tridiagonal component matrices in $\mathcal{O}(N)$ computational complexity.
Utilized the extended Clausius-Clapeyron equation ($P^s = \exp(C_1 + C_2/T + C_3\ln T + C_4 T^{C_5})$) for temperature-dependent vapor pressures and stage bubble-point updates.
Extended the column model to 10 equilibrium stages and conducted feed tray location sensitivity analysis to minimize thermal fluctuations and optimize separation purity.
Implemented rigorous unit test suites with unittest in test.py covering matrix tridiagonal solvers, enthalpy balances, and convergence routines.

2. Machine Learning & Deep Learning for Thermophysical Property Estimation

Domain: Scientific Machine Learning & Molecular Thermodynamics
Primary Report: 22CH30008_PMS_A1_Report.pdf (Section 2)
Jupyter Notebook: Sol_A1.ipynb
Visualizations: fig1.png, fig2.png, fig4.png

Problem Summary:

Estimating critical thermophysical properties (normal boiling point $T_b$ and reduced boiling point $T_b/T_c$) of over 600 organic chemical compounds from molecular weight ($M_w$) and Pitzer acentric factor ($\omega$).

Key Technical Contributions:

Baseline Parametric Regression: Derived simple and multivariable linear regressions. Formulated closed-form Ordinary Least Squares via the Normal Equation: $$\boldsymbol{\theta} = (\mathbf{X}^T \mathbf{X})^{-1} \mathbf{X}^T \mathbf{y}$$ Achieved $R^2 = 0.7861$ and $MSE = 0.0004$ on reduced boiling point correlation ($\theta_0 = 0.5958$, $\theta_1 = 0.0002$, $\theta_2 = 0.1546$).
Cost Function Landscape Visualization: Rendered 3D meshgrid contour plots of the Mean Squared Error loss surface against parameter vectors to evaluate convexity and gradient behavior.
Deep Neural Network (MLP): Designed a deep feedforward regression network in TensorFlow/Keras trained under extreme data scarcity (10% training, 90% test set).
Hyperparameter Optimization & TensorBoard: Executed grid search across network depths (1–4 hidden layers) and widths (16–64 neurons/layer) with ReLU activations and Adam optimization.
Results: Best architecture (3 hidden layers, 16 neurons/layer) attained a test MSE of $0.00057$ and MAE of $0.0166$ ($< 1.7%$ error), demonstrating superior non-linear property representation over traditional empirical correlations.

3. Non-Linear Multicomponent Flash Separation & Recovery Optimization

Domain: Non-Linear Thermodynamics & Equilibrium Optimization
Primary Report: 22CH30008_PMS_A1_Report.pdf (Section 3)
Source Code: B_a_b.py, B_c.py

Problem Summary:

Separation of a ternary aromatic mixture (100 kmol/h feed: 60 mol% Benzene, 25 mol% Toluene, 15 mol% o-Xylene) in an isothermal flash drum operating at 1.01325 bar across temperatures from 100°C to 150°C.

Key Technical Contributions:

Formulated the non-linear Rachford-Rice objective function: $$f(\psi) = \sum_{i=1}^C \frac{z_i (K_i - 1)}{1 + \psi (K_i - 1)} = 0$$ and derived its exact analytical first derivative for second-order quadratic convergence via Newton-Raphson iteration.
Dynamically integrated Antoine vapor pressure equations to update equilibrium ratios $K_i(T)$.
Base Case ($100^\circ\text{C}$): Converged in 4 iterations to vapor fraction $\psi = 0.6840$ ($V = 68.4\text{ kmol/h}$, $L = 31.6\text{ kmol/h}$) with vapor phase composition $y = [0.6963, 0.2245, 0.0793]$.
Process Optimization: Conducted fine-grid temperature sweeps ($0.0001^\circ\text{C}$ resolution) to maximize Benzene recovery in distillate, identifying the global optimum at $T = 106.99^\circ\text{C}$ yielding a peak recovery of $92.47%$.

4. Debutanizer Column Design via Fenske-Underwood-Gilliland (FUG) Shortcut

Domain: Hydrocarbon Fractionation & Shortcut Column Design
Primary Report: 22CH30008.pdf (Part 1)
Source Code: 1.py, test.ipynb

Problem Summary:

Design and specification of an 8-component petroleum debutanizer column ($i\text{C}_4$, $n\text{C}_4$, $i\text{C}_5$, $n\text{C}_5$, $\text{C}_6$, $\text{C}_7$, $\text{C}_8$, $\text{C}_9$) operating at 80 psia (552 kPa) to fractionate Light Key ($n\text{C}_4$) and Heavy Key ($i\text{C}_5$).

Key Technical Contributions:

Fenske Equation: Estimated minimum theoretical stages ($N_{min} = 14.93$) and calculated product splits for all 6 non-key components across top and bottoms streams.
Underwood Equation: Numerically solved the non-linear root condition using scipy.optimize.fsolve for the active root $\theta$ between relative volatilities to evaluate the minimum reflux ratio ($R_{min} = 91.77$).
Gilliland Correlation: Automated calculation of actual theoretical stages required at an operating reflux ratio of $1.3 R_{min}$, determining $N = 27.26$ equilibrium stages.
Validated hydrocarbon K-values under elevated pressure and evaluated operational feasibility.

5. Pinch Analysis, Temperature Cascade & Heat Exchanger Network (HEN) Synthesis

Domain: Process Integration, Energy Efficiency & Thermal Pinch Analysis
Primary Reports:
- Pinch Analysis and Heat Exchanger Network Integration.docx
- 22CH30008.pdf (Part 2)
- PMS A2-1.pdf, A2 Q3 PMS.pdf
Visuals & Hand Synthesis: 3 graph.jpeg

Problem Summary:

Energy integration of a multi-stream chemical process network consisting of 3 hot streams and 3 cold streams with a minimum temperature approach $\Delta T_{min} = 10^\circ\text{C}$.

Key Technical Contributions:

Constructed Composite Curves (Hot and Cold curves on a $T\text{-}H$ diagram) to visualize maximum heat exchange potential.
Formulated the Problem Table Algorithm (Temperature Cascade Method) using shifted interval temperatures ($T^* = T - \Delta T_{min}/2$ for hot streams, $T^* = T + \Delta T_{min}/2$ for cold streams) to identify enthalpy surpluses and deficits.
Pinch Point Determination: Identified the exact pinch point at cold stream $90^\circ\text{C}$ / hot stream $100^\circ\text{C}$.
Utility Targets: Evaluated minimum hot utility demand ($Q_{H,min} = 0.1\text{ MW}$), minimum cold utility demand ($Q_{C,min} = 0.05\text{ MW}$), and recovered $1.8\text{ MW}$ of heat internally.
HEN Synthesis & Flowsheet Integration: Designed an optimal heat exchanger network respecting pinch heuristics (no heat transfer across pinch, no cold utilities above pinch, no hot utilities below pinch) and synthesized a full Process Flow Diagram (PFD) with labeled heat duties and split streams.

6. Industrial Shell-and-Tube Heat Exchanger Rating & Retrofit (Aspen HYSYS & EDR)

Domain: Computer-Aided Process Engineering (CAPE) & Industrial Thermal Design
Primary Report: Assignment 2 CAPE LAb report.pdf
Simulation Files: cape lab.hsc, cape lab.bk0, A2Q1.hsc, A2Q1.bk0

Problem Summary:

Thermal design, rating, and retrofitting of an industrial shell-and-tube heat exchanger in Aspen HYSYS and Aspen EDR (Exchanger Design & Rating).

Key Technical Contributions:

Base Case Design: Simulated a multi-pass (2-pass tube, 1-pass shell) exchanger heating 80,000 lb/hr Benzene (70°F to 140°F, 45 psia) using Toluene (235°F to 150°F, 40 psia) with allowable $\Delta P \le 5\text{ psia}$ and fouling resistance $R_f = 0.0015\text{ ft}^2\cdot\text{hr}\cdot^\circ\text{F/BTU}$. Calculated total heat duty of $2,124,640\text{ BTU/hr}$ ($2.125\text{ MMBTU/hr}$).
Detailed EDR Rating: Selected TEMA shell type, baffle cuts, tube bundle pitch, diameter, and metallurgy to satisfy pressure drop constraints and heat transfer coefficients.
Retrofit / Re-Rating Scenario: Evaluated the existing physical unit for a severe new process duty: heating Methanol using high-temperature pressurized water at $10.5\text{ MMBTU/hr}$ (a ~5x duty increase). Retuned operating pressures, stream allocations, and checked overdesign/vibration margins.

7. Multicomponent Reboiled Absorber Simulation (Simultaneous Correction Method)

Domain: Complex Separation Processes & Simultaneous Non-Linear Solvers
Primary Code: 2.py
Problem Definition: A2.pdf (Question 2)

Problem Summary:

Rigorous simulation of a 13-stage deethanizing reboiled absorber column operating at 400 psia (2.76 MPa) with n-decane absorbent oil separating light hydrocarbons ($\text{C}_1, \text{C}_2, \text{C}_3, n\text{C}_4, n\text{C}_5$).

Key Technical Contributions:

Implemented the Simultaneous Correction (SC) algorithm to simultaneously solve coupled stage material balances, phase equilibria, and non-linear enthalpy balances.
Interfaced CoolProp thermodynamic library (PropsSI) for cryogenic and high-pressure liquid/vapor hydrocarbon enthalpies.
Iteratively converged stage temperatures, interstage flow rates, and reboiler duty under high pressure.

Mathematical Formulations & Algorithms

1. Rachford-Rice Equation (Flash VLE)

The objective function for vapor fraction $\psi = V/F$: $$f(\psi) = \sum_{i=1}^C \frac{z_i (K_i - 1)}{1 + \psi(K_i - 1)} = 0$$ Analytical derivative for Newton-Raphson update: $$f'(\psi) = - \sum_{i=1}^C \frac{z_i (K_i - 1)^2}{[1 + \psi(K_i - 1)]^2}$$ Newton-Raphson iteration step: $$\psi^{(k+1)} = \psi^{(k)} - \frac{f(\psi^{(k)})}{f'(\psi^{(k)})}$$

2. Thomas Algorithm (Tridiagonal Matrix Algorithm - TDMA)

For stage-by-stage component mass balance tridiagonal systems: $$A_j x_{j-1} + B_j x_j + C_j x_{j+1} = D_j$$ Forward elimination: $$C'j = \frac{C_j}{B_j - A_j C'{j-1}}, \quad D'j = \frac{D_j - A_j D'{j-1}}{B_j - A_j C'_{j-1}}$$ Back substitution: $$x_j = D'_j - C'j x{j+1}$$

3. Normal Equation (Linear Regression)

$$\boldsymbol{\theta} = (\mathbf{X}^T \mathbf{X})^{-1} \mathbf{X}^T \mathbf{y}$$ $$\text{Cost: } J(\boldsymbol{\theta}) = \frac{1}{2m} \sum_{i=1}^m (\mathbf{x}^{(i)\boldsymbol{\theta}} - y^{(i)})^2$$

4. Fenske-Underwood-Gilliland (FUG) Equations

Fenske Equation (Minimum theoretical stages): $$N_{min} = \frac{\ln\left[\left(\frac{x_{LK,D}}{x_{HK,D}}\right)\left(\frac{x_{HK,B}}{x_{LK,B}}\right)\right]}{\ln \alpha_{LK/HK}}$$
Underwood Equation (Minimum reflux ratio): $$\sum_{i=1}^C \frac{\alpha_i z_{i,F}}{\alpha_i - \theta} = 1 - q, \quad R_{min} = \sum_{i=1}^C \frac{\alpha_i x_{i,D}}{\alpha_i - \theta} - 1$$

Repository Structure

Assigmnet PMS/
├── README.md                                          # Comprehensive project documentation & CV guide
├── PMS A2-1.pdf                                       # Assignment 2 Question 3 handwritten solution & HEN
│
├── Assignment 1/                                      # ASSIGNMENT 1: ML, Flash, & Bubble Point Distillation
│   ├── A1.pdf                                         # Original assignment problem sheet
│   ├── 22CH30008_PMS_A1_Report.pdf                    # Complete compiled engineering report (12 pages)
│   ├── data_file.xlsx                                 # Thermophysical properties dataset (600+ compounds)
│   ├── test.py                                        # Unit testing suite for distillation module
│   ├── Sol_A!_q3.ipynb                                # Distillation prototyping notebook
│   ├── Problem1/                                      # Q1: Regression & Neural Networks
│   │   ├── Sol_A1.ipynb                               # End-to-end ML notebook (EDA, OLS, ANN, TensorBoard)
│   │   └── fig1.png, fig2.png, fig4.png               # Regression fits, 3D loss surface & residual plots
│   ├── Problem2/                                      # Q2: Multicomponent Flash Problem
│   │   ├── B_a_b.py                                   # Newton-Raphson Rachford-Rice solver & temperature sweep
│   │   └── B_c.py                                     # Benzene recovery optimization algorithm
│   └── Problem3/                                      # Q3: Multicomponent Distillation (Wang-Henke BP)
│       ├── C_a hand iteration (1).pdf                 # Step-by-step verified hand calculation
│       ├── C_b.py                                     # OOP Distillation class with Thomas Algorithm
│       └── C_c.py                                     # 10-stage distillation & feed location optimizer
│
├── Assignment 2/                                      # ASSIGNMENT 2: FUG, SC Method, & Pinch Analysis
│   ├── A2.pdf                                         # Original assignment problem sheet
│   ├── 22CH30008.pdf                                  # Debutanizer & Pinch Analysis engineering report
│   ├── Pinch Analysis and Heat Exchanger Network...docx # Comprehensive HEN design & cascade report
│   ├── 2.py                                           # Simultaneous Correction (SC) Reboiled Absorber
│   ├── test.ipynb                                     # FUG shortcut distillation calculations
│   ├── Problem 1/                                     # Q1: FUG Shortcut Distillation
│   │   └── 1.py                                       # Fenske, Underwood, & Gilliland Python solver
│   └── Problem 3/                                     # Q3: Pinch Analysis & Flowsheet Synthesis
│       ├── A2 Q3 PMS.pdf                              # Detailed handwritten cascade & HEN calculations
│       └── 3 graph.jpeg                               # Composite curves & heat exchanger grid diagram
│
└── CAPE Lab/                                          # COMPUTER-AIDED PROCESS ENGINEERING (ASPEN)
    ├── Assignment 2 CAPE LAb report.pdf               # Complete Aspen HYSYS/EDR rating report (10 pages)
    ├── cape lab.hsc, cape lab.bk0                     # Aspen HYSYS simulation case & backup files
    ├── A2Q1.hsc, A2Q1.bk0                             # Heat exchanger simulation files
    ├── Assignment 2/                                  # HYSYS output stream and block property reports
    │   ├── q1p1.txt, q1p2.txt, q1p3.txt, q2p1.txt
    └── arpit/                                         # Aspen Plus simulation models (.apwz)

Installation & Execution Guide

Prerequisites

Python 3.8 to 3.10
Jupyter Notebook / JupyterLab
Commercial license for Aspen HYSYS & EDR (for .hsc and .bk0 files)

1. Clone Repository & Setup Environment

git clone https://github.com/arpitkumar2004/Assigmnet-PMS.git
cd Assigmnet-PMS

# Create virtual environment
python -m venv venv
# Activate on Windows:
.\venv\Scripts\activate
# Activate on Linux/macOS:
source venv/bin/activate

# Install required packages
pip install numpy scipy pandas matplotlib seaborn tensorflow scikit-learn openpyxl coolprop

2. Running Simulations

Run Multicomponent Distillation Unit Tests:

cd "Assignment 1"
python test.py

Run Multicomponent Flash & Recovery Optimization:

cd "Assignment 1/Problem2"
python B_c.py

Run Debutanizer FUG Shortcut Calculations:

cd "Assignment 2/Problem 1"
python 1.py

Run Reboiled Absorber Simultaneous Correction Solver:

cd "Assignment 2"
python 2.py

Launch Jupyter Notebooks:

jupyter notebook
# Open Assignment 1/Problem1/Sol_A1.ipynb for ML/ANN workflows
# Open Assignment 2/test.ipynb for FUG distillation workflows

Author

Arpit Kumar
Department of Chemical Engineering, Indian Institute of Technology (IIT) Kharagpur
GitHub: @arpitkumar2004

Name		Name	Last commit message	Last commit date
.vscode		.vscode
A3 - sol		A3 - sol
Assignment 1		Assignment 1
Assignment 2		Assignment 2
CAPE Lab		CAPE Lab
__pycache__		__pycache__
a1Bpms		a1Bpms
logs		logs
README.md		README.md

Latest commit

History

Folders and files

Repository files navigation

Process Modelling, Simulation & Computer-Aided Process Engineering (PMS & CAPE)

Table of Contents

Portfolio Overview & CV Highlights

Core Competencies Demonstrated:

Important Project Reports (Ready for CV/Portfolio)

1. Multicomponent Multistage Distillation Simulation (Wang-Henke BP Method)

Problem Summary:

Key Technical Contributions:

2. Machine Learning & Deep Learning for Thermophysical Property Estimation

Problem Summary:

Key Technical Contributions:

3. Non-Linear Multicomponent Flash Separation & Recovery Optimization

Problem Summary:

Key Technical Contributions:

4. Debutanizer Column Design via Fenske-Underwood-Gilliland (FUG) Shortcut

Problem Summary:

Key Technical Contributions:

5. Pinch Analysis, Temperature Cascade & Heat Exchanger Network (HEN) Synthesis

Problem Summary:

Key Technical Contributions:

6. Industrial Shell-and-Tube Heat Exchanger Rating & Retrofit (Aspen HYSYS & EDR)

Problem Summary:

Key Technical Contributions:

7. Multicomponent Reboiled Absorber Simulation (Simultaneous Correction Method)

Problem Summary:

Key Technical Contributions:

Mathematical Formulations & Algorithms

1. Rachford-Rice Equation (Flash VLE)

2. Thomas Algorithm (Tridiagonal Matrix Algorithm - TDMA)

3. Normal Equation (Linear Regression)

4. Fenske-Underwood-Gilliland (FUG) Equations

Repository Structure

Installation & Execution Guide

Prerequisites

1. Clone Repository & Setup Environment

2. Running Simulations

Run Multicomponent Distillation Unit Tests:

Run Multicomponent Flash & Recovery Optimization:

Run Debutanizer FUG Shortcut Calculations:

Run Reboiled Absorber Simultaneous Correction Solver:

Launch Jupyter Notebooks:

Author

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages