Assignment – Optimization of chemical processes

1. The total annualized cost for a counter-flow heat exchanger is the sum of its capital cost (Ccap∝A0.8)
and operating cost (Cop∝A−0.2ΔT−0.5, where A is area). Formulate the objective function to minimize
total cost. Apply the principles of any optimization techniques to explain why this posynomial
structure is ideally suited for this method and what non-linear insights the technique provides before
any numerical solution.

2. A chemical plant must supply five process heaters with steam from three different sources, each
with a different cost per ton and limited capacity. Formulate this system as a Linear Programming (LP)
problem to minimize the total steam cost. Interpret the physical and economic meaning of the Dual
Variables (Shadow Prices) associated with the heater demand constraints, and Evaluate how this
information can justify investing in a more efficient heat recovery system for the highest-cost steam
user.

3. The optimal time to chemically clean a heat exchanger is found by minimizing the total cost function
C(t)=Cclean/t+Cenergy(t), where Cenergy(t) is the non-linear energy penalty due to fouling. The first
derivative, ∂C/∂t, has been derived. Apply the Newton-Raphson (or Secant) Method to find the
optimal cleaning interval (t∗). Critically Evaluate why the Secant Method might be preferred over the
full Newton's method in this practical engineering application, even though Newton's method
generally converges faster.

4. The total annualized cost of a distillation column is a unimodal function of the reflux ratio (R), which
is expensive to simulate. Design an optimization strategy using a Region Elimination Method (e.g.,
Golden Section Search or Polynomial Interpolation) to find the optimal reflux ratio within 1%
uncertainty in the minimum number of simulation runs. Justify why this direct search method is
suitable for a unimodal function where the analytical derivative is complex or unavailable.

5. You need to separate four components (A, B, C, D) using a sequence of three binary distillation
columns. There are multiple possible sequences (e.g., direct-sequence vs. indirect-sequence). Apply
the principle of any Programming to systematically determine the optimal separation sequence that
minimizes the total sum of energy costs. Illustrate how the 'Principle of Optimality' allows you to
break this large, multi-stage decision problem into a series of smaller, manageable subproblems.

6. A new pipeline route requires selecting the optimal pipe diameter (D) from a set of standard
available pipe sizes (e.g., 6", 8", 10", 12"). The objective is to minimize the sum of the capital cost
(Cpipe∝D) and the pumping cost (Cpump∝D−5). Formulate this selection process as an Integer
Programming (IP) problem (Unit III) using binary variables. Justify why relaxing this problem to a
continuous Non-Linear Program and rounding the result might lead to a suboptimal or even physically
infeasible solution.

7. You are modeling the annualized cost of a pumping system which is a function of the flow rate (Q)
and the pump efficiency (η). The resulting cost function is found to be complex and non-convex.
Classify the resulting objective function (e.g., Convex/Non-convex, Multimodal/Unimodal). Analyze
the critical implications of a non-convex (multimodal) cost surface on finding the global optimum,
particularly when using gradient-based optimization solvers that rely solely on the Necessary and
Sufficient Conditions for Optimum.

8. The cost of a reactor system is a non-linear, quadratic function of the required reactor volume (V).
The volume, however, must satisfy several linear inequality constraints related to maximum heat
removal capacity, minimum catalyst loading, and available space. Formulate this system as a
Quadratic Programming (QP) problem . Analyze why the QP formulation is often preferred over
general Non-Linear Programming for problems with a quadratic objective function and linear
constraints in terms of computational speed and guaranteed finding of the global minimum.

9. For an exothermic reaction system, the relationship between reaction temperature (T) and reaction
rate (r) is often a multimodal function with a minimum rate followed by a maximum rate. Critically
Evaluate the physical significance of the two stationary points identified by the Necessary Condition
for Optimum (∂r/∂T=0). Determine which stationary point corresponds to the optimal operating
temperature for maximum throughput and which must be avoided for stable operation, using the
concept of the Sufficient Condition for Optimum (i.e., the sign of the second derivative).

10. A multi-product plant must determine its daily production schedule (amount of P1,P2,P3) to
maximize total profit, subject to constraints on raw materials and minimum customer demand for
product P3. The minimum demand constraint necessitates the introduction of a surplus variable.
Formulate the problem and Explain why the Big-M Method or the Two-Phase Simplex Method (Unit
II) would likely be required to find an initial basic feasible solution. Illustrate the role of the artificial
variables in this process.

Part B

1. A simplified model for the operating cost (C) of a separator as a function of its temperature (T) is
given by C(T)=0.01T2−4T+500. Apply the Necessary Condition for Optimum to find the stationary
point T∗. Apply the Sufficient Condition for Optimum to determine if this point is a local minimum or
a local maximum.

2. A designer wants to minimize the sum of two solvent flow rates, S1+S2, subject to the constraint
that their product must be 100 (S1S2=100). Set up the Lagrangian function and describe the steps you
would take to use the Lagrange Multiplier method to find the optimal flow rates.

3. A simple distillation column produces two products, PA and PB, requiring limited amounts of steam
and cooling water. After formulating and solving the LP, the Simplex Tableau shows slack variables for
both steam and water have zero values in the final solution. Analyze the physical meaning of these
zero slack values in terms of resource utilization (i.e., are the steam and water constraints active?).

4. A pump is transferring fluid between two tanks. The optimization goal is to minimize the annual
cost of the fluid transport system. Identify two primary design variables (that you can select) and two
primary fixed parameters (that you cannot change) that would typically be included in the cost
objective function.

5. A process requires installing a single valve. Three standardized valve sizes (V1,V2,V3) are available,
and the goal is to minimize cost. Formulate a simple Integer Programming constraint using binary
variables (yi) that enforces the requirement: "Exactly one of the three valves must be selected."

6. A batch reactor's total profit is a function of the reaction time (t) and appears to have a local
maximum at t=2 hours and a higher global maximum at t=8 hours. Sketch this multimodal profit
function and explain why a pure analytical method relying on derivatives might fail to find the global
optimum t=8 if the starting guess is close to t=2.