Thermal System Design, Optimization, and Decision-Making

Introduction and Context

Heat and mass transfer equipment rarely operates as an isolated component. A heat exchanger is connected to pumps, compressors, control valves, piping, utilities, sensors, and production equipment. Its design therefore affects energy consumption, product quality, safety, maintenance, environmental performance, and financial results. A technically excellent design may be unacceptable if it is difficult to manufacture, unreliable in service, hazardous to maintain, or uneconomic over its life.

Thermal system design is the structured process of converting a required duty into a safe, manufacturable, controllable, maintainable, and economically justified system. It combines conservation laws and constitutive models with engineering judgment and decision analysis. The designer must consider not only the design-point temperature and flow rate, but also start-up, shutdown, fouling, off-design operation, uncertainty, degradation, and future operating conditions.

This chapter develops an integrated framework for such decisions. It emphasizes the connection between heat- and mass-transfer calculations and broader design criteria: material selection, manufacturability, reliability, maintenance, lifecycle cost, exergy, energy efficiency, risk, and multi-criteria optimization.

Learning Objectives

After studying this chapter, a student should be able to:

1. Formulate thermal-system design requirements using energy and mass balances.
2. Select and interpret heat-transfer and mass-transfer performance models.
3. Compare materials using thermal, mechanical, chemical, manufacturing, and lifecycle criteria.
4. Distinguish energy efficiency from exergy efficiency.
5. Estimate lifecycle cost and include maintenance, downtime, and replacement effects.
6. Apply reliability, risk, and uncertainty concepts to thermal equipment.
7. Formulate single-objective and multi-objective optimization problems.
8. Make and defend a design recommendation using transparent engineering criteria.

1. From Thermal Duty to System Requirements

A design begins with a duty statement, not with a particular heat exchanger or material. The duty statement should specify:

• process streams and mass-flow rates;
• inlet and required outlet temperatures or concentrations;
• allowable pressure drop;
• operating and design pressures;
• fluid composition, phase, toxicity, corrosiveness, and fouling tendency;
• operating schedule and expected lifetime;
• available utilities and utility costs;
• required availability, safety level, and regulatory constraints.

For a steady-flow control volume, the general energy balance is

where ṁ is mass-flow rate, h is specific enthalpy, V is velocity, g is gravitational acceleration, z is elevation, Q̇ is heat transfer to the control volume, and Ẇ is work delivered by it. For many heat exchangers, changes in kinetic and potential energy are negligible and shaft work is absent, giving approximately

for a single stream. If a single-phase fluid has approximately constant specific heat,

where c_p is specific heat capacity in J·kg⁻¹·K⁻¹. This relation is applicable when phase change is absent and variations in c_p are small over the temperature range.

For a two-stream exchanger with negligible heat loss to the surroundings,

The equality is a conservation requirement, not a sizing equation. It must be satisfied before area, geometry, or material is selected.

Mass transfer likewise begins with a species balance. For steady operation,

where Ṅ_i represents generation or consumption of species i. In an absorber, for example, the decrease in solute flow in the gas must equal the increase in solute flow in the liquid, subject to reaction and accumulation assumptions.

2. Thermal and Mass-Transfer Models for Design

2.1 Overall heat-transfer coefficient

For a heat exchanger, the basic design relation is

where U is the overall heat-transfer coefficient, A is the selected heat-transfer area, F is a correction factor for multipass or cross-flow arrangements, and ∆T_lm is the logarithmic mean temperature difference (LMTD). For counterflow or parallel flow,

Here ∆T₁ and ∆T₂ are the temperature differences at the two ends, defined consistently according to the flow arrangement. The LMTD method is most useful when all terminal temperatures are known. If outlet temperatures are unknown and heat-capacity rates are known, the effectiveness–NTU method is often more suitable.

The overall coefficient is controlled by a series of thermal resistances. For a cylindrical wall, a representative resistance expression is

where h_i and h_o are inside and outside convection coefficients, R_f,i and R_f,o are fouling resistances, k_w is wall thermal conductivity, r_i and r_o are wall radii, and L is length. The areas used in each term must be consistent with the chosen overall coefficient basis. This equation assumes one-dimensional radial conduction, steady operation, negligible contact resistance, and constant properties.

A frequent design error is to increase wall conductivity when convection or fouling dominates the resistance. Resistance analysis identifies where improvement will have the greatest effect.

2.2 Convection, conduction, and pressure drop

Convection correlations are commonly expressed using dimensionless groups:

Nu is the Nusselt number, Re the Reynolds number, and Pr the Prandtl number. Their use is valid only within the geometry, flow regime, property range, and boundary conditions for which the correlation was developed.

Pressure drop is not a secondary issue. Pumping or compression power may be approximated by

where ∆p is pressure drop, Q̇_v is volumetric flow rate, and η_p is pump efficiency. In a complete evaluation, the increased heat-transfer area obtained from higher velocity must be weighed against the increased operating power, erosion, vibration, and noise.

2.3 Mass-transfer design

For a dilute species transferred across a gas–liquid interface, a practical flux model is

where N_A is molar transfer rate per volume or area according to the definition of K_y a, K_y is an overall gas-side coefficient, a is interfacial area per packed volume, y_A is the bulk gas concentration expressed as a fraction, and y_A^* is the equilibrium value corresponding to the liquid phase. The model assumes local equilibrium at the interface and a suitable dilute or averaged-property approximation.

Mass-transfer equipment must also be checked for flooding, entrainment, pressure drop, foaming, solvent degradation, and chemical compatibility. A mathematically adequate transfer coefficient does not guarantee a robust industrial design.

3. Material Selection as a Systems Decision

Material selection should not be reduced to thermal conductivity. Relevant criteria include:

The material must satisfy constraints before it is ranked. For example, a low-cost alloy that fails a pressure-vessel code requirement is infeasible, regardless of its high score on cost.

Material properties are temperature- and environment-dependent. Yield strength, fatigue strength, thermal conductivity, expansion coefficient, and corrosion rate may change substantially during operation. Differential thermal expansion can create stresses at joints. A high-conductivity material may nevertheless be inappropriate if it causes galvanic corrosion or cannot be welded reliably.

Manufacturability includes tolerances, surface finish, tube expansion, brazing or welding procedures, heat treatment, nondestructive examination, and access for inspection. Design for manufacture and assembly (DFMA) encourages fewer parts, standardized joints, accessible fasteners, and processes supported by qualified suppliers.

4. Energy, Exergy, and Environmental Performance

Energy is conserved, but its usefulness is degraded by irreversibility. First-law efficiency measures quantity; exergy analysis measures potential to produce useful work relative to an environment.

For a system receiving useful energy output Ė_useful and energy input Ė_in, energy efficiency is

Exergy efficiency is

where Eẋ_product is the exergy associated with the desired result and Eẋ_fuel is the exergy supplied. Exergy destruction is related to entropy generation by the Gouy–Stodola relation:

where T₀ is ambient absolute temperature and Ṡ_gen is entropy-generation rate. The relation applies to an adequately defined control volume and a specified reference environment.

A heat exchanger can transfer a large amount of energy while destroying substantial exergy if heat crosses a large temperature difference. Exergy analysis may therefore favor heat recovery, temperature matching, reduced mixing, and lower approach temperatures. However, reducing temperature differences usually requires more area and capital. The decision is economic and operational, not purely thermodynamic.

5. Reliability, Maintenance, and Risk

Reliability is the probability that equipment performs its required function for a specified time under stated conditions. For a constant failure rate λ, the reliability function is

This exponential model is a simplification and is most appropriate for a useful-life period with an approximately constant hazard rate. Real equipment may show early failures, random failures, and wear-out failures.

Availability includes repair time. A simple steady-state approximation is

where MTBF is mean time between failures and MTTR is mean time to repair. Availability is improved by reliability, rapid diagnosis, spare parts, access, isolation valves, bypasses, and maintainable layouts.

Risk is commonly represented conceptually as

This expression is useful for prioritization but does not capture every dimension of uncertainty. A formal hazard and operability study (HAZOP), failure mode and effects analysis (FMEA), or fault-tree analysis may be required. Typical thermal-system hazards include overpressure, loss of cooling, flammable release, toxic exposure, hot surfaces, vacuum collapse, thermal runaway, and water hammer.

Safety factors must not substitute for hazard identification. Compliance with pressure-vessel, fire, environmental, electrical, and occupational-safety regulations is a design constraint, not an optional optimization criterion.

6. Lifecycle Cost and Economic Evaluation

Initial purchase cost is only one part of total ownership cost. A lifecycle model may include capital cost, installation, energy, water, chemicals, labor, scheduled maintenance, unplanned downtime, replacement, disposal, and residual value.

The present value of a future cost C_t occurring at year t is

where i is the discount rate. The net present cost over n years is

where C₀ is initial cost and S_t is salvage value or other credit. All monetary quantities must use the same price basis and currency convention. Inflation and discounting should not be mixed incorrectly; either use nominal cash flows with a nominal discount rate or real cash flows with a real discount rate.

A design with higher capital cost may be preferable if it reduces fouling, pumping power, failures, or production losses. Sensitivity analysis should test uncertain electricity price, utilization, lifetime, fouling rate, and discount rate.

7. Optimization and Decision Frameworks

A single-objective design can be written as

where x is the vector of design variables, J is the objective, g_j are inequality constraints, and h_k are equality constraints. Variables may include area, diameter, flow velocity, material, number of passes, control settings, insulation thickness, and maintenance interval.

Typical objectives are lifecycle cost, exergy destruction, energy use, mass, emissions, or risk. They often conflict. Increasing area can reduce utility consumption but increase capital cost. Increasing velocity can increase heat transfer but also pressure drop and erosion.

A multi-criteria method may use a weighted score:

where z_r are normalized criterion scores and w_r are stakeholder weights. Scores must be normalized so that “higher is better” has the same interpretation for every criterion. Weight selection should be documented and tested through sensitivity analysis.

A Pareto-optimal design is one for which no objective can be improved without worsening at least one other objective. The Pareto frontier is often more informative than a single “best” answer because it reveals trade-offs to decision-makers.

Uncertainty should be propagated rather than hidden. Monte Carlo simulation can vary uncertain parameters such as fouling coefficient, demand, utility price, and failure rate. Robust design seeks acceptable performance across a range of conditions, not merely optimum performance at one nominal point.

8. Worked Example: Preliminary Exchanger Decision

A process stream of water must be heated from 25 °C to 65 °C at ṁ = 2 kg/s. Take c_p = 4.18 kJ·kg⁻¹·K⁻¹. The required duty is

Assume a preliminary overall coefficient of U = 800 W·m⁻²·K⁻¹, a counterflow LMTD of 30 K, and F = 1. The required clean area is

This is a preliminary area, not a final purchase specification. If fouling reduces the effective coefficient to 600 W·m⁻²·K⁻¹, the corresponding area becomes

The additional 4.65 m² may be justified by reduced cleaning frequency and improved availability. The decision must also check velocity, pressure drop, tube vibration, material compatibility, thermal expansion, code requirements, and cleaning access. The example illustrates why design margin should be connected to a failure mechanism rather than added arbitrarily.

9. Illustrative Industrial Case: Waste-Heat Recovery System

Background

A food-processing plant discharges a warm, dilute wastewater stream and uses natural gas to generate hot water. Management is considering a stainless-steel plate heat exchanger, a polymer-lined exchanger, or continued direct boiler heating.

Problem

The recovery system must lower fuel consumption without causing unacceptable fouling, hygiene risk, production downtime, or difficult maintenance. Wastewater composition varies seasonally, and cleaning requires a scheduled production stoppage.

Analysis

The engineering team first measures flow, temperature, suspended solids, pH, and fouling behavior. A heat balance estimates recoverable duty. Pressure-drop calculations establish pump requirements. A fouling resistance is included in the exchanger model rather than using only clean performance. Materials are screened for food-contact compatibility, corrosion resistance, gasket temperature limits, and cleanability.

The alternatives are evaluated against capital cost, annual fuel savings, electricity use, cleaning labor, expected availability, hygienic risk, and greenhouse-gas emissions. A discounted lifecycle model is used over ten years. Failure scenarios include leakage between streams, loss of flow, blocked channels, gasket failure, and inadequate cleaning.

Approach and solution

The selected design uses a removable plate exchanger with differential-pressure monitoring, temperature sensors on both streams, isolation valves, a bypass, and a clean-in-place connection. The design includes spare gaskets and a documented inspection interval. A control system reduces heat recovery when approach temperature or pressure drop indicates fouling. The decision is based on a Pareto comparison: a larger exchanger has higher capital cost but lower operating temperature difference and greater fouling tolerance.

Outcome

In an illustrative evaluation, the removable exchanger provides the best balance of fuel savings and maintainability, while the lowest-cost alternative has unacceptable cleaning labor and contamination risk. The final decision is conditional on pilot fouling tests and verification of hygienic design requirements.

Learning points

1. Heat recovery must be designed around actual contaminants and operating variability.
2. Maintainability can have greater economic value than a small increase in clean thermal performance.
3. Instrumentation provides both control and early warning of degradation.
4. A lifecycle and risk comparison is more defensible than selecting solely on purchase price.

10. Digital Tools, AI, and Contemporary Practice

Computer-aided design, computational fluid dynamics, process simulation, digital twins, and optimization software can accelerate design. However, a model is only as reliable as its boundary conditions, property data, calibration, and validation. CFD should not replace experimental checks when turbulence, multiphase flow, fouling, or conjugate heat transfer is uncertain.

Machine-learning models can predict fouling, detect abnormal temperature profiles, and estimate remaining useful life from sensor data. Their deployment requires representative data, sensor-quality checks, explainability appropriate to the risk, cybersecurity, and human oversight. An AI recommendation should be treated as decision support, not as evidence that physical constraints or safety obligations have been satisfied.

Contemporary thermal design increasingly addresses decarbonization, electrification, heat pumps, industrial heat recovery, low-global-warming refrigerants, additive manufacturing of compact exchangers, and advanced coatings. These technologies introduce new questions concerning supply chains, repairability, material recycling, fire safety, and long-term evidence.

11. Best-Practice Design Workflow

A disciplined workflow is:

1. Define duty, boundaries, stakeholders, and regulatory requirements.
2. Establish mass and energy balances.
3. Select validated heat- and mass-transfer correlations.
4. Identify design variables, constraints, and uncertain parameters.
5. Screen materials for compatibility, strength, manufacturability, and availability.
6. Size equipment at clean and degraded conditions.
7. Estimate pressure drop, utility demand, exergy destruction, and emissions.
8. Perform reliability, maintainability, and hazard analyses.
9. Calculate lifecycle cost and conduct sensitivity analysis.
10. Compare feasible alternatives using transparent criteria.
11. Verify through testing, inspection, commissioning, and performance monitoring.
12. Update the design basis using operating data and lessons learned.

Challenges and Limitations

Design correlations may not represent unusual geometries or dirty, transient, multiphase flows. Property data may be uncertain, and material degradation may be nonlinear. Economic forecasts and failure probabilities are inherently uncertain. Weighted decision matrices can conceal stakeholder disagreement or create false precision. Exergy results depend on the selected reference environment. Optimization algorithms may converge to mathematically attractive but physically impractical solutions. Finally, lifecycle benefits may be undermined by poor installation, inadequate operator training, or unavailable spare parts.

Good engineering therefore combines calculation with validation, conservative interpretation of uncertainty, independent review, and clear documentation of assumptions.

Chapter Summary

Thermal system design is an integrated decision process rather than a heat-transfer calculation alone. Conservation laws establish required duty, while convection, conduction, fouling, and mass-transfer models determine equipment performance. Material selection must include corrosion, strength, compatibility, joining, inspection, availability, and lifecycle effects. Energy efficiency measures quantity of energy, whereas exergy analysis identifies degradation of useful energy and the location of irreversibility.

Reliability, availability, maintainability, and risk determine whether equipment can deliver its duty safely over time. Lifecycle costing includes capital, utilities, maintenance, downtime, replacement, and disposal. Optimization must respect physical, safety, manufacturing, and regulatory constraints. Because cost, energy, risk, environmental impact, and performance conflict, Pareto analysis, weighted criteria, sensitivity analysis, and robust design are valuable decision tools. Digital technologies and AI can support monitoring and optimization, but they require validated models, reliable data, cybersecurity, and human accountability.

Key Terms

Thermal system design: Integrated design of equipment and supporting systems to deliver a required thermal function.

Overall heat-transfer coefficient: Combined coefficient representing series thermal resistances.

Fouling resistance: Additional thermal resistance caused by deposits or surface contamination.

LMTD: Logarithmic mean temperature difference used in exchanger sizing.

Effectiveness: Actual heat transfer divided by the maximum possible heat transfer.

NTU: Number of transfer units, a dimensionless measure of exchanger size and conductance.

Mass-transfer coefficient: Parameter relating species flux to a concentration or potential difference.

Manufacturability: Degree to which a design can be produced, joined, inspected, and repaired.

DFMA: Design for manufacture and assembly.

Energy efficiency: Useful energy output divided by energy input.

Exergy: Maximum useful work obtainable relative to a reference environment.

Exergy destruction: Loss of work potential caused by irreversibility.

Reliability: Probability of performing a required function for a specified time.

Availability: Fraction of time equipment is capable of performing its function.

MTBF: Mean time between failures.

MTTR: Mean time to repair.

Risk: Combination of likelihood and consequence of an unwanted event.

Lifecycle cost: Present-value cost over acquisition, operation, maintenance, and disposal.

Pareto optimum: Design for which improvement in one objective requires deterioration in another.

Robust design: Design that maintains acceptable performance despite uncertainty and variation.

Review Questions

Short Answer Questions

1. Why is a duty statement required before selecting a heat exchanger?
2. What physical effects are represented by fouling resistance?
3. Distinguish energy efficiency from exergy efficiency.
4. Define MTBF and MTTR and explain their relationship to availability.
5. Why should material selection include manufacturability?

Long Answer Questions

1. Explain how the thermal-resistance network influences exchanger improvement strategies.
2. Describe a lifecycle-cost methodology for comparing two thermal-system designs.
3. Explain how reliability, maintainability, and risk analysis influence equipment configuration.
4. Discuss the advantages and limitations of weighted decision matrices and Pareto analysis.
5. Describe a complete workflow for designing a heat-recovery system under uncertain fouling conditions.

Analytical / Application Questions

1. A stream of 1.5 kg/s is heated by 35 K with specific heat 4.0 kJ·kg⁻¹·K⁻¹. Calculate the required heat-transfer rate and identify two additional data items needed for exchanger sizing.
2. Two designs have equal duty, but Design A has lower capital cost and higher pressure drop, while Design B has higher capital cost and lower pressure drop. Formulate the lifecycle and technical comparisons needed before choosing one.
3. A plant observes a gradual increase in exchanger pressure drop and a decrease in outlet temperature. Propose a diagnostic plan using measurements, likely failure mechanisms, and a maintenance decision rule.

Multiple Choice Questions with Answers

Question 1

For a single-phase stream with constant specific heat, which relation estimates its heat-transfer rate?

A. Q̇ = ṁ c_p ∆T

B. Q̇ = U/A

C. Q̇ = ρV/μ

D. Q̇ = T₀ Ṡ_gen

Correct Answer: A. Q̇ = ṁ c_p ∆T

Question 2

Which resistance is added to an exchanger model to represent deposits on a heat-transfer surface?

A. Contact resistance only

B. Fouling resistance

C. Radiation resistance

D. Inertial resistance

Correct Answer: B. Fouling resistance

Question 3

Exergy destruction is related to entropy generation by which expression?

A. Eẋ_destroyed = Ṡ_gen/T₀

B. Eẋ_destroyed = T₀ Ṡ_gen

C. Eẋ_destroyed = T₀/Ṡ_gen

D. Eẋ_destroyed = Q̇/T₀

Correct Answer: B. Eẋ_destroyed = T₀ Ṡ_gen

Question 4

A pump handles Q̇_v = 0.020 m³/s against ∆p = 100 kPa at efficiency 0.80. What is its approximate power requirement?

A. 0.4 kW

B. 1.6 kW

C. 2.5 kW

D. 4.0 kW

Correct Answer: C. 2.5 kW

Question 5

Which criterion is a lifecycle consideration rather than only a clean thermal-performance criterion?

A. Clean overall heat-transfer coefficient

B. Logarithmic mean temperature difference

C. Expected cleaning and downtime cost

D. Terminal temperature difference

Correct Answer: C. Expected cleaning and downtime cost

Question 6

For a constant failure rate model, reliability varies with time as:

A. R(t)=1−e^(−λt)

B. R(t)=e^(−λt)

C. R(t)=λt

D. R(t)=1/(λt)

Correct Answer: B. R(t)=e^(−λt)

Question 7

A Pareto-optimal design is one for which:

A. All objectives have their individual unconstrained minima

B. No objective can improve without worsening at least one other objective

C. Capital cost is always minimized

D. All design variables are equal

Correct Answer: B. No objective can improve without worsening at least one other objective

Question 8

Which action most directly improves maintainability?

A. Removing isolation valves

B. Increasing operating temperature without analysis

C. Providing access, bypasses, and replaceable components

D. Using an unvalidated correlation

Correct Answer: C. Providing access, bypasses, and replaceable components

Question 9

If a future cost of 10,000 monetary units occurs in year 2 at a discount rate of 5%, its present value is approximately:

A. 9,070

B. 9,524

C. 10,500

D. 11,025

Correct Answer: B. 9,524

Question 10

Why should an AI prediction of exchanger fouling be treated as decision support?

A. AI models automatically satisfy pressure-vessel codes

B. Physical constraints, data quality, and validation remain necessary

C. AI eliminates the need for sensors

D. AI guarantees zero maintenance

Correct Answer: B. Physical constraints, data quality, and validation remain necessary