Skip to content

Lesson 1: Operations Research and the Modelling Approach

1.1 Learning Outcomes

By the end of this hour a student will be able to:

  1. Define operations research and recount its origins.
  2. List and explain the phases of an OR study.
  3. Classify mathematical models by type and choose an appropriate class.
  4. Identify decision variables, the objective function and constraints in a verbal problem.
  5. Formulate a simple real-world problem as a mathematical programme.

1.2 What Is Operations Research?

Definition. Operations Research (OR) is the application of scientific and mathematical methods to the study of the operations of a system, with the aim of providing decision-makers with an optimal or near-optimal course of action.

Alternative definitions worth knowing.

  • (Churchman, Ackoff and Arnoff) "OR is the application of scientific methods, techniques and tools to problems involving the operations of systems, so as to provide those in control of the system with optimum solutions to the problems."
  • (Operational Research Society, UK) "The attack of modern science on complex problems arising in the direction and management of large systems of men, machines, materials and money."

The three defining characteristics.

  1. A systems orientation. OR examines the whole system, not an isolated part. Optimising one department may harm the organisation — a phenomenon called sub-optimisation.
  2. An interdisciplinary team. Mathematicians, statisticians, engineers, economists and domain experts working together.
  3. A scientific method. Observe, formulate a hypothesis (model), test, validate, implement.

Historical origins

Period Development
1937–1945 Born in Britain during WWII. Teams studied radar deployment, convoy sizing and anti-submarine tactics. The term "operational research" dates from Bawdsey Research Station, 1937.
1940s Blackett's Circus (P. M. S. Blackett's team) — the archetypal interdisciplinary OR group.
1947 George Dantzig invents the simplex method — the single most consequential event in the field.
1950s OR moves into industry: transportation, inventory, queueing. Journals and societies founded.
1950s–60s Dynamic programming (Bellman), network flows (Ford–Fulkerson), PERT/CPM (1958).
1979 Khachiyan's ellipsoid method — LP is proved to be in P.
1984 Karmarkar's interior-point method — polynomial and practical.
1990s–present Massive solver improvements (CPLEX, Gurobi); OR merges with machine learning and data science.

A striking fact about progress. Between 1990 and 2020, LP solvers improved by roughly \(10^{6}\)× from algorithms and \(10^{6}\)× from hardware — a combined \(10^{12}\)× speed-up. A problem that took a year in 1990 now takes about 30 milliseconds.

Classic wartime example — the convoy problem

The question. Should merchant ships cross the Atlantic in small frequent convoys or large infrequent ones?

The intuition (wrong): small convoys are less conspicuous.

The OR analysis. A convoy's escort ring scales with its perimeter (\(\propto\sqrt{n}\) for \(n\) ships in a compact formation), while the ships protected scale with \(n\). So escort strength per ship improves as convoys grow. Furthermore, U-boat detection probability depends on convoy frequency, not size.

The conclusion: large convoys. Implementing this cut shipping losses substantially, and it is the canonical illustration of OR overturning plausible intuition with quantitative reasoning.

1.3 Phases of an OR Study

   1. Formulate the problem
              |
   2. Construct the model
              |
   3. Derive a solution      <---+
              |                  |
   4. Test / validate           | (iterate)
              |                  |
   5. Establish controls  ------+
              |
   6. Implement

Phase 1 — Formulating the problem. Identify the decision-maker, the objectives, the alternatives available and the constraints. Typically the hardest and most neglected phase. A precisely solved wrong problem is worthless.

Phase 2 — Constructing the model. Translate the verbal problem into mathematics: decision variables, objective function, constraints. Decide what to simplify.

Phase 3 — Deriving a solution. Apply an algorithm — simplex, branch and bound, a queueing formula, simulation.

Phase 4 — Testing and validation. Does the model reproduce known past behaviour? Are the answers sensible under extreme inputs? Retrospective testing against historical data is standard.

Phase 5 — Establishing controls. The solution is valid only while the parameters hold. Sensitivity analysis identifies which parameters matter and how far they may drift.

Phase 6 — Implementation. Translate the mathematical solution into operating procedures people will actually follow. More OR studies fail here than at any technical stage.

1.4 Types of Models

By degree of abstraction

Type Description Example
Iconic a scaled physical replica a wind-tunnel aircraft model
Analogue one property represents another a contour map; an electrical circuit modelling a mechanical system
Symbolic / mathematical variables and equations \(\max\ 3x_1+5x_2\) subject to constraints

OR works almost entirely with symbolic models.

By certainty

Type Description Techniques
Deterministic all parameters known with certainty LP, transportation, assignment, CPM, EOQ
Probabilistic / stochastic some parameters are random variables queueing, PERT, stochastic inventory, Markov chains
Game-theoretic an intelligent opponent chooses adversarially game theory

By mathematical structure

Model Objective Constraints Variables
Linear programming linear linear continuous
Integer programming linear linear integer
Mixed integer linear linear both
Non-linear programming non-linear any continuous
Quadratic programming quadratic linear continuous
Dynamic programming any staged any

This course covers all six.

1.5 Anatomy of a Mathematical Programme

Every optimisation model has exactly three components.

\[ \begin{aligned} \text{Optimise } \quad & Z = f(x_1, x_2, \ldots, x_n) && \text{\textbf{objective function}}\\ \text{subject to } \quad & g_i(x_1,\ldots,x_n)\ \{\le,\ =,\ \ge\}\ b_i, \quad i = 1,\ldots,m && \text{\textbf{constraints}}\\ & x_j \ge 0, \quad j = 1,\ldots,n && \text{\textbf{non-negativity}} \end{aligned} \]

1. Decision variables. The quantities under the decision-maker's control. Always define them with units and a time base: "\(x_1\) = tonnes of product A produced per week", not "\(x_1\) = product A".

2. Objective function. The single quantity to maximise (profit, throughput) or minimise (cost, time, distance).

3. Constraints. Limitations: resource availability, demand requirements, capacity, policy rules, and non-negativity.

Terminology

Term Meaning
Feasible solution any \(\vec x\) satisfying all constraints
Feasible region the set of all feasible solutions
Optimal solution a feasible solution giving the best objective value
Infeasible no solution satisfies all constraints
Unbounded the objective can be improved without limit
Redundant constraint removing it does not change the feasible region
Binding (active) constraint holds with equality at the optimum
Slack the unused amount of a \(\le\) resource
Surplus the excess over a \(\ge\) requirement

1.6 Solved Problems

Problem 1 — Product-mix formulation

A furniture firm makes tables and chairs. A table needs 5 hours of carpentry and 2 hours of finishing; a chair needs 3 hours of carpentry and 4 hours of finishing. Weekly capacity is 120 hours of carpentry and 90 hours of finishing. Profit is ₹450 per table and ₹300 per chair. Formulate an LP to maximise weekly profit.

Solution.

Step 1 — Decision variables.

\[ x_1 = \text{number of tables produced per week} \]
\[ x_2 = \text{number of chairs produced per week} \]

Step 2 — Objective function. Maximise weekly profit (in ₹):

\[ \max\ Z = 450x_1 + 300x_2 \]

Step 3 — Constraints.

Carpentry: each table uses 5 h, each chair 3 h; 120 h available.

\[ 5x_1 + 3x_2 \le 120 \]

Finishing: each table 2 h, each chair 4 h; 90 h available.

\[ 2x_1 + 4x_2 \le 90 \]

Non-negativity: you cannot produce a negative quantity.

\[ x_1,\ x_2 \ge 0 \]

The complete model.

\[ \begin{aligned} \max\ & Z = 450x_1 + 300x_2\\ \text{s.t. } & 5x_1 + 3x_2 \le 120 && \text{(carpentry)}\\ & 2x_1 + 4x_2 \le 90 && \text{(finishing)}\\ & x_1, x_2 \ge 0 \end{aligned} \]

A useful presentation device — the data table. Always tabulate before formulating:

Table (\(x_1\)) Chair (\(x_2\)) Available
Carpentry (h) 5 3 120
Finishing (h) 2 4 90
Profit (₹) 450 300 maximise

Reading off the model. Each row is a constraint; the bottom row is the objective. Students who build this table first make far fewer formulation errors.

(This problem is solved graphically in L03 and by simplex in L05; the optimum is \(x_1 = 15\), \(x_2 = 15\), \(Z = ₹11{,}250\).)

Problem 2 — Diet problem formulation

A nutritionist must design a minimum-cost daily diet from two foods. Food A costs ₹12 per unit and provides 4 units of protein, 2 of fat and 8 of carbohydrate. Food B costs ₹20 per unit and provides 9 units of protein, 5 of fat and 3 of carbohydrate. The diet must supply at least 36 units of protein and 20 of fat, and at most 40 units of carbohydrate. Formulate.

Solution.

Decision variables.

\[ x_1 = \text{units of Food A per day}, \qquad x_2 = \text{units of Food B per day} \]

Objective. Minimise cost:

\[ \min\ Z = 12x_1 + 20x_2 \]

Constraints.

\[ \begin{aligned} 4x_1 + 9x_2 &\ge 36 && \text{(protein — at least)}\\ 2x_1 + 5x_2 &\ge 20 && \text{(fat — at least)}\\ 8x_1 + 3x_2 &\le 40 && \text{(carbohydrate — at most)}\\ x_1, x_2 &\ge 0 \end{aligned} \]

Note the mixed inequality directions. Nutritional requirements give \(\ge\); limits give \(\le\). Confusing them is the most common formulation error in diet problems. A reliable check: ask "would more of this be a problem or a benefit?" — benefit ⟹ \(\ge\); problem ⟹ \(\le\).

Historical note. The diet problem was posed by the US Army in 1945 and solved by George Stigler using heuristics (77 foods, 9 nutrients); Dantzig later solved it exactly by simplex in 1947 — one of the first practical LP applications.

Problem 3 — Blending problem with a ratio constraint

A refinery blends two crude oils into petrol. Crude 1 costs ₹60/barrel and is 40 % octane; Crude 2 costs ₹85/barrel and is 70 % octane. At least 8000 barrels of blend are needed, with an octane content of at least 55 %. Supply of Crude 2 is limited to 5000 barrels. Formulate.

Solution.

Decision variables.

\[ x_1 = \text{barrels of Crude 1},\qquad x_2 = \text{barrels of Crude 2} \]

Objective.

\[ \min\ Z = 60x_1 + 85x_2 \]

Constraint 1 — total volume.

\[ x_1 + x_2 \ge 8000 \]

Constraint 2 — octane content. The blend's octane fraction must be at least 0.55:

\[ \frac{0.40x_1 + 0.70x_2}{x_1 + x_2} \ \ge\ 0.55 \]

This is non-linear as written. Multiply through by \((x_1+x_2) > 0\) — legitimate because the denominator is strictly positive:

\[ 0.40x_1 + 0.70x_2 \ \ge\ 0.55x_1 + 0.55x_2 \]
\[ -0.15x_1 + 0.15x_2 \ \ge\ 0 \qquad\Longrightarrow\qquad \boxed{\ x_2 \ \ge\ x_1\ } \]

Now linear. The blend needs at least as much high-octane crude as low-octane crude.

Constraint 3 — supply limit.

\[ x_2 \le 5000 \]

Complete model.

\[ \begin{aligned} \min\ & Z = 60x_1 + 85x_2\\ \text{s.t. } & x_1 + x_2 \ge 8000\\ & x_2 - x_1 \ge 0\\ & x_2 \le 5000\\ & x_1, x_2 \ge 0 \end{aligned} \]

The lesson. Ratio and percentage requirements are always linearisable by cross-multiplying, provided the denominator is guaranteed positive. This trick appears constantly in blending, portfolio and staffing problems, and it is worth memorising.

(Quick feasibility check: \(x_1 + x_2 \ge 8000\) with \(x_2\le5000\) and \(x_2\ge x_1\) forces \(x_1 \ge 3000\) and \(x_1\le x_2\le 5000\) — the region is non-empty ✓.)

Problem 4 — Multi-period production planning

A firm faces demand of 100, 150 and 120 units in months 1, 2 and 3. Production capacity is 130 units per month. Production costs ₹50 per unit; holding inventory costs ₹8 per unit per month. Initial inventory is zero. Formulate an LP minimising total cost.

Solution.

Decision variables — note that two families are needed:

\[ p_t = \text{units produced in month } t, \qquad t = 1,2,3 \]
\[ I_t = \text{inventory carried from month } t \text{ to } t+1, \qquad t = 1,2,3 \]

Objective.

\[ \min\ Z = 50(p_1+p_2+p_3) + 8(I_1+I_2+I_3) \]

Constraints — inventory balance. This is the key structural idea:

\[ \boxed{\ \text{opening inventory} + \text{production} = \text{demand} + \text{closing inventory}\ } \]
\[ \begin{aligned} \text{Month 1:}\quad & 0 + p_1 = 100 + I_1 &&\Rightarrow\ p_1 - I_1 = 100\\ \text{Month 2:}\quad & I_1 + p_2 = 150 + I_2 &&\Rightarrow\ I_1 + p_2 - I_2 = 150\\ \text{Month 3:}\quad & I_2 + p_3 = 120 + I_3 &&\Rightarrow\ I_2 + p_3 - I_3 = 120 \end{aligned} \]

Capacity.

\[ p_t \le 130, \qquad t = 1,2,3 \]

Non-negativity. \(p_t, I_t \ge 0\). (Requiring \(I_t \ge 0\) is what forbids backordering; allowing \(I_t\) to go negative would model backorders.)

Why production must be smoothed. Total demand is \(100+150+120 = 370\); total capacity is \(3\times130 = 390\). But month 2's demand of 150 exceeds the 130 monthly capacity, so the firm must build inventory in month 1.

The obvious feasible plan. Produce \(p_1 = 120\) (carrying \(I_1 = 20\)), \(p_2 = 130\) (carrying \(I_2 = 0\)), \(p_3 = 120\):

\[ Z = 50(370) + 8(20 + 0 + 0) = 18{,}500 + 160 = ₹18{,}660 \]

Since production cost is constant per unit and total production is fixed at 370, minimising cost means minimising inventory, and carrying exactly 20 units for one month is the minimum required. So this plan is optimal, with \(Z = ₹18{,}660\).

(If production costs varied by month — overtime, seasonal labour — the trade-off would become genuinely interesting, and this is the standard lot-sizing model.)

Problem 5 — Recognising sub-optimisation

A manufacturer's production department is rewarded for minimising unit cost, and therefore runs very long production batches. The warehouse is separately rewarded for minimising inventory. Sales is rewarded for never missing an order. Explain the conflict in OR terms and state the correct formulation.

Solution.

The conflict.

Department Local objective Local optimum Effect elsewhere
Production minimise unit cost very large batches (setup cost spread thin) huge inventory
Warehouse minimise inventory very small batches frequent setups, high cost
Sales never stock out large safety stock high inventory

Each department is behaving rationally given its own metric, and the firm as a whole suffers. This is sub-optimisation: optimising components separately does not optimise the system.

Why it happens mathematically. Let \(Q\) be the batch size, \(D\) annual demand, \(C_o\) setup cost and \(C_h\) holding cost per unit per year. The two local objectives are

\[ \text{Production:}\ \min\ \frac{DC_o}{Q} \ \Rightarrow\ Q\to\infty \]
\[ \text{Warehouse:}\ \min\ \frac{QC_h}{2} \ \Rightarrow\ Q\to 0 \]

They pull in opposite directions, and neither alone has an interior optimum.

The correct (system-wide) formulation. Minimise the sum:

\[ \min_{Q>0}\ TC(Q) = \frac{DC_o}{Q} + \frac{QC_h}{2} \]

Differentiating and setting to zero:

\[ \frac{dTC}{dQ} = -\frac{DC_o}{Q^{2}} + \frac{C_h}{2} = 0 \qquad\Longrightarrow\qquad \boxed{\ Q^{*} = \sqrt{\frac{2DC_o}{C_h}}\ } \]

This is the Economic Order Quantity, derived in full in L31. Note that it sits between the two departmental extremes, and neither department would ever propose it.

The organisational fix. Change the incentives, not just the mathematics: reward departments on total system cost, or use transfer prices that internalise the externality. This is why OR implementation (Phase 6) so often requires organisational change, and why so many technically correct OR studies fail.

\[ \boxed{\ \textbf{OR's first principle: optimise the system, not its parts.}\ } \]

1.7 Practice Problems

  1. A company makes two products needing 2 and 3 hours of machine time, with 60 hours available; profits are ₹40 and ₹50. Formulate.
  2. Formulate a diet problem with three foods and two nutrients, choosing your own data.
  3. A blend must be at most 30 % impurity from two sources with 20 % and 45 % impurity. Write the linear constraint. [Ans: \(-0.10x_1 + 0.15x_2 \le 0\), i.e. \(3x_2 \le 2x_1\)]
  4. Extend Problem 4 to 4 months with your own demand figures.
  5. Give two more examples of sub-optimisation from everyday organisations.
  6. Classify each as deterministic or stochastic: EOQ, PERT, simplex, M/M/1, CPM.
  7. State the six phases of an OR study and which is most often the cause of failure.

1.8 Summary

  • OR applies scientific method to system-level decision-making; its three hallmarks are a systems orientation, an interdisciplinary team and a scientific method.
  • Born in WWII Britain; transformed by Dantzig's simplex method (1947); LP solvers have improved by roughly \(10^{12}\)× since 1990.
  • Six phases: formulate → model → solve → validate → control → implement. Formulation and implementation, not solution, are where studies usually fail.
  • Models are iconic / analogue / symbolic, and deterministic / stochastic / game-theoretic.
  • Every mathematical programme has decision variables, an objective function and constraints; define variables with units and a time base.
  • Ratio and percentage requirements linearise by cross-multiplication when the denominator is positive.
  • Multi-period models are built on the inventory-balance identity: opening + production = demand + closing.
  • Sub-optimisation — optimising parts separately — is OR's central cautionary principle; the EOQ formula is the classic resolution.

Next lesson: L02 — Introduction to linear programming and formulation.

Practise now: open Interactive-Web-Platform/unit1-lp-lab.html → Model Builder: enter a verbal problem's data in a table and watch the LP formulation assemble itself.