Production planning and control

Mechanical Engineering (B.E./B.Tech) · Industrial engineering · free study notes for exams, placements and interviews

In short

Production planning and control (PPC) decides what to make, how much, on which machine, in what order and by when, then follows the shop floor to see that it actually happens. Planning converts a demand forecast into a master schedule, a capacity plan, routes and loads; control converts the schedule into dispatched work orders and corrects deviations through expediting. It is a heavily examined area in Industrial Engineering papers and in GATE ME (forecasting, line balancing, sequencing, scheduling) and is the daily job of planning engineers at Maruti Suzuki, Tata Motors, Ashok Leyland, Bosch, BHEL, HAL and L&T. Interviewers test it because it links design, manufacturing, costing and delivery performance in one decision.

Explained

What PPC is: objectives and functions

Definition: the direction and coordination of a firm's men, materials and machines so that the right quantity of the right quality is produced at the right time at minimum cost.
Objectives: meet delivery dates (on-time-in-full), maximise utilisation of plant and labour, minimise work-in-process (WIP) and lead time, minimise cost, and keep production level enough to avoid hiring-firing churn.
Pre-planning functions: forecasting, product and quantity decision, process planning (what operations, in what sequence), routing (which machines/work-centres and the path through the shop), estimating (standard times from work study, costs), loading (how much work is put on each work-centre), scheduling (calendar dates and times).
Action function: dispatching — releasing the authority, drawings, materials, tools and job cards to start work.
Control functions: expediting / progress chasing / follow-up, comparison of actual against plan, corrective action (overtime, subcontracting, re-routing, re-scheduling), and reporting.
Pre-requisites of good PPC: sound forecast, reliable bill of materials and routes, realistic standard times, stable supplier deliveries, honest shop-floor feedback.

Production systems and planning horizons

Job (jobbing) production: one-off or very small lots, general-purpose machines, process (functional) layout, high skill, high variety, long and variable lead time — die shops, L&T heavy fabrication, repair shops. Scheduling is the hardest here.
Batch production: lots of identical items repeated periodically; setup cost drives the batch size; machine groups or cell layout — pumps, valves, pharmaceuticals, garments.
Mass (flow) production: one or few products, product (line) layout, special-purpose machines, work moves by conveyor, PPC reduces to line balancing and material feeding — car assembly at Manesar, two-wheeler assembly at Hosur.
Continuous process production: uninterrupted flow, 24×7 operation, plant is the process — Bhilai Steel Plant, cement, refineries, fertiliser. Here PPC is campaign planning and shutdown planning.
Order policies: MTS (make-to-stock, forecast-driven, FMCG, spares), MTO (make-to-order, customer order drives production, machine tools), ATO (assemble-to-order from standard modules, PCs, commercial vehicle variants), ETO (engineer-to-order, turnkey plant).
Planning hierarchy: long range (1-5 years: plant, product mix, capital capacity) → medium range (3-18 months: aggregate production plan) → short range (weeks: MPS, MRP, loading) → very short (daily/hourly: dispatching, sequencing).
Push vs pull: PPC schedules releases from a forecast (push); a pull system authorises production only on downstream consumption (treated in the lean note). Most Indian plants run a hybrid — MPS at the top, kanban between adjacent stages.

Forecasting demand

Qualitative (new product, no history): Delphi method (iterative anonymous expert rounds), market survey, sales-force composite, executive opinion, historical analogy.
Quantitative time-series: simple moving average (stable demand, lags trend), weighted moving average, simple exponential smoothing (one constant α, weights decay geometrically), trend-adjusted / double exponential smoothing (Holt) for a trend, seasonal indices (ratio-to-moving-average) for repeating peaks, least-squares linear trend.
Causal / associative: linear regression of demand on an independent variable (tractor demand on monsoon rainfall, cement on construction index).
Choice of α: small α (0.05-0.2) smooths noise but lags change; large α (0.3-0.5) responds fast but chases noise.
Error measures: bias (RSFE), MAD, MSE, MAPE. A tracking signal = RSFE/MAD outside about ±4 means the model is biased and must be revised (trend or seasonality present).
Rules that matter in exams and practice: forecast error falls as you aggregate (family level is more accurate than SKU level) and rises with horizon; forecast the family, disaggregate by mix ratios.

Capacity planning, aggregate planning and the master schedule

Capacity types: *design/rated* capacity (ideal), *effective* capacity (after allowances for setup, maintenance, breaks), *actual* output. Utilisation = actual/rated; efficiency = actual/effective.
Capacity requirement is computed from quantity × standard time, inflated for scrap/yield, and compared with available machine-hours and man-hours; the gap is closed by overtime, extra shift, subcontracting or new machines.
Aggregate production planning (APP): plans output, inventory and workforce in aggregate units (equivalent units, machine-hours) for 3-18 months.
- Level strategy: constant output; absorbs demand swings with inventory and backorders — low labour cost, high carrying cost.
- Chase (matching) strategy: output follows demand by hiring/firing or varying hours — low inventory, high hiring/overtime cost and poor morale.
- Mixed strategy: stable core workforce + overtime, temporary/contract labour and subcontracting — what Indian plants actually use.
- Costs traded off: regular time, overtime (double wages by statute), hiring/training, layoff, inventory carrying, stockout/backorder, subcontracting.
Master production schedule (MPS): a time-phased statement of *end items* to be built in each period, within the aggregate plan; checked by rough-cut capacity planning (RCCP) against key work-centres before release.
Available-to-promise (ATP) = on-hand + scheduled receipts − committed orders; it is what sales may promise a customer.
The MPS explodes through the bill of materials into component and purchase requirements — the MRP mechanics belong to the inventory/MRP note; here it is enough that MPS stability (a "frozen" near horizon) is what keeps MRP from nervousness.
Capacity requirement planning (CRP) then converts planned orders into load profiles per work-centre; overloads are levelled by shifting orders.

Routing, loading, scheduling and sequencing

Routing output documents: route sheet / operation sheet (operation sequence, machine, tool, jig, standard time), material requisition, job card / time ticket, move ticket, inspection ticket.
Loading: infinite loading ignores capacity and shows the resulting overload; finite loading assigns only up to available capacity; forward scheduling starts from today (earliest finish, builds WIP), backward scheduling starts from the due date and works back (used by MRP, minimises inventory but no slack for upsets).
Gantt chart: bar chart of jobs against time per machine; shows load, progress and idleness. Variants: load chart, progress chart. Still the shop-floor display of choice.
Priority (dispatching) rules for a single machine / shop:
- FCFS — fair, simple, poor performance.
- SPT (shortest processing time) — minimises mean flow time, mean lateness and average WIP; long jobs starve.
- EDD (earliest due date) — minimises maximum lateness/tardiness.
- LPT — used to load parallel machines early.
- Slack = due date − today − remaining processing time; MINSLACK or slack-per-operation.
- Critical ratio CR = (due date − today)/remaining processing time: CR < 1 behind schedule, = 1 on schedule, > 1 ahead; job with the lowest CR goes first. Dynamic, so it is recomputed each time.
- Moore's algorithm minimises the *number* of tardy jobs.
Flow-shop sequencing: Johnson's rule gives the minimum-makespan sequence for *n* jobs on 2 machines in the same order — pick the smallest time; if it is on machine 1 place that job as early as possible, if on machine 2 as late as possible; delete and repeat. Extension to 3 machines is valid only if min t1 ≥ max t2 or min t3 ≥ max t2; then combine A = t1 + t2, B = t2 + t3 and apply Johnson. Job-shop scheduling with m machines is NP-hard — solved by heuristics, priority rules or APS software.
Performance measures: makespan (total elapsed time), machine idle time, mean flow time, lateness (can be negative), tardiness (never negative), number of tardy jobs, WIP.
Little's law links the three: average WIP = throughput rate × average flow time — the quickest argument for cutting batch sizes and queues.

Line balancing and bottlenecks

Purpose: divide total assembly work into stations so that each station's time ≤ cycle time, with as few stations and as little idle time as possible, respecting the precedence diagram.
Cycle time is set by demand: Cmax = available time per shift / units required (this is takt time when set by customer demand).
Steps: list tasks and times → draw precedence diagram → compute cycle time → compute theoretical minimum stations → assign tasks by a heuristic → compute efficiency.
Heuristics: Largest Candidate Rule (longest eligible task first), Ranked Positional Weight (Helgeson-Birnie) (rank by task time plus all followers), Kilbridge-Wester (assign by precedence column).
Measures: line efficiency, balance delay (= 1 − efficiency), smoothness index. Output rate is fixed by the slowest station, not the average.
Mixed-model lines (several variants on one line, as in car assembly) need model sequencing to level the work content and part consumption; two-sided and U-shaped lines are used to cut walking and station count.
Theory of constraints: identify the bottleneck → exploit it (never let it idle or run scrap) → subordinate everything else → elevate it → repeat. Schedule the bottleneck first and protect it with a buffer (drum-buffer-rope); an hour lost at the bottleneck is an hour lost by the whole plant.

Dispatching, control and current practice

Dispatching issues the order; centralised dispatching plans centrally for all shops (better overall view), decentralised lets each foreman sequence within the shop (faster response).
Control loop: plan → release → record actual (machine logs, SCADA, barcode/RFID, MES terminals) → compare → corrective action. Reports: production status, delayed-order, machine-utilisation, scrap/rework, shortage ("red-line") list.
Expediting is triggered by material shortage, tool/fixture non-availability, breakdown, absenteeism, rework or a design change.
Shop-floor realities built into plans: setup times and setup-reduction, machine availability from the maintenance plan, operator learning on new work (learning curve), rejection/rework yields, shift pattern.
Statutory limits that shape Indian schedules: the Occupational Safety, Health and Working Conditions Code, 2020 was brought into force on 21 November 2025 and subsumes the Factories Act, 1948. It caps adult work at 8 h/day and 48 h/week, and allows overtime only with the worker's consent and at twice the ordinary rate of wages; rest interval, spread-over and the overtime ceiling per quarter are fixed by central/state rules, which differ, so confirm the current state notification before planning overtime. (The Factories Act figures still quoted in textbooks: 9 h/day, 48 h/week, 10½ h spread-over, half an hour of rest after 5 hours, one weekly holiday, overtime at double wages.)
Systems: ERP modules (SAP PP/PP-DS, Oracle, Microsoft Dynamics) for MPS/MRP/CRP; MES for shop-floor execution; APS for finite-capacity scheduling; Industry 4.0 additions — IoT machine data, digital twin, OEE dashboards, AI demand forecasting. The planning logic examined here is exactly the logic these packages implement.

Key points

PPC - planning (routing, loading, scheduling) plus control (dispatching, expediting, corrective action).
Routing - the path and sequence of operations; scheduling - when each will be done; loading - how much work a work-centre carries.
MTS vs MTO - stock made against forecast versus production triggered by a customer order; ATO and ETO lie in between.
Aggregate plan - 3-18 month output/workforce/inventory plan in aggregate units; level, chase or mixed strategy.
MPS - time-phased end-item build plan, validated by rough-cut capacity planning, frozen in the near horizon.
Utilisation = actual/rated capacity, efficiency = actual/effective capacity; never quote one for the other.
Exponential smoothing - F(new) = F(old) + α(error); lags a trend, so check the tracking signal.
MAD, MAPE, tracking signal - accuracy, percentage accuracy and bias of a forecast.
SPT minimises mean flow time; EDD minimises maximum tardiness; CR is the dynamic rule used on the floor.
Johnson's rule - minimum makespan for n jobs on 2 machines in the same order; 3 machines only under its special condition.
Makespan vs flow time - total elapsed time for the set versus average time a job spends in the system.
Tardiness ≥ 0, lateness can be negative - an early job has negative lateness but zero tardiness.
Cycle time = available time / required output; the slowest station, not the average, fixes line output.
Balance delay = 1 − line efficiency = idle time fraction of total station time.
Little's law - WIP = throughput × flow time; shorter queues directly cut lead time.
Bottleneck governs throughput - schedule it first, buffer it, never starve it.
Forward scheduling builds WIP and slack; backward scheduling from the due date minimises inventory but no cushion.
Working-hours law - OSH&WC Code, 2020 (in force 21 Nov 2025, replacing the Factories Act, 1948): 8 h/day, 48 h/week, overtime at double wages; the limit on any shift plan.

Formulas and examples

Simple moving average: F(t+1) = (D(t) + D(t−1) + … + D(t−n+1)) / n — n = periods averaged
Weighted MA: F(t+1) = Σ wᵢ Dᵢ , Σ wᵢ = 1
Exponential smoothing: F(t+1) = α D(t) + (1−α) F(t) = F(t) + α e(t); 0 < α < 1, e(t) = D(t) − F(t)
Trend-adjusted (Holt): L(t) = α D(t) + (1−α)(L(t−1)+T(t−1)); T(t) = β(L(t)−L(t−1)) + (1−β)T(t−1); F(t+k) = L(t) + kT(t)
Seasonal index SI = average demand of the season / overall average demand; forecast = deseasonalised trend × SI
MAD = Σ|e| / n; MSE = Σe²/n; MAPE = (100/n) Σ|e/D| %; RSFE = Σe; Tracking signal = RSFE / MAD (limits ≈ ±4)
Utilisation = actual output / rated capacity; Efficiency = actual output / effective capacity
Input needed for one stage = required good output / (1 − scrap fraction); multi-stage: divide by the product of yields
Machines required N = (Q × t) / (H × E), Q = pieces to be made, t = time per piece, H = hours available per machine, E = availability × efficiency
Economic batch quantity EBQ = √[2 D Cs / (Ch (1 − d/p))] (batch sizing; the inventory note develops it)
Cycle time Cmax = available production time per period / units required; Takt time = same, set by customer demand
Theoretical minimum stations Nmin = ⌈Σ tᵢ / Cmax⌉ (round UP)
Line efficiency η = Σ tᵢ / (N × Cmax); Balance delay = 1 − η; Smoothness index = √[Σ (Cmax − tₛ)²]
Flow time Fᵢ = completion time Cᵢ − release time; Mean flow time = ΣCᵢ / n
Lateness Lᵢ = Cᵢ − dᵢ; Tardiness Tᵢ = max(0, Cᵢ − dᵢ); Makespan = time from start of first job to finish of last
Critical ratio CR = (due date − present date) / remaining processing time; Slack = due date − present date − remaining processing time
Little's law: WIP = throughput rate × average flow time
Learning curve: T(n) = T(1) × n^b, b = log r / log 2 (r = learning rate, e.g. 0.8 for an 80% curve)

Example 1 - Exponential smoothing with an error check. Demand (units/month): 1,100; 1,050; 1,200; 1,150. Forecast for month 1 was 1,000, α = 0.2.

F2 = 1,000 + 0.2(1,100 − 1,000) = 1,020
F3 = 1,020 + 0.2(1,050 − 1,020) = 1,026
F4 = 1,026 + 0.2(1,200 − 1,026) = 1,026 + 34.8 = 1,060.8
F5 = 1,060.8 + 0.2(1,150 − 1,060.8) = 1,060.8 + 17.84 ≈ 1,079 units
Errors: +100, +30, +174, +89.2 → MAD = 393.2/4 = 98.3 units; MAPE = (0.0909 + 0.0286 + 0.1450 + 0.0776)/4 × 100 = 8.6%
RSFE = +393.2, tracking signal = 393.2/98.3 = +4.0 — every error is positive, so the model is lagging a rising trend; switch to a trend-adjusted model or raise α.

Example 2 - Machine requirement and utilisation. 60,000 good gear blanks a year; machining time 4.5 min/piece; scrap 4%; plant works 300 days, two 8-hour shifts; machine availability × efficiency = 0.85.

Input to be machined = 60,000 / 0.96 = 62,500 pieces
Machine time needed = 62,500 × 4.5 = 281,250 min = 4,687.5 h
Effective hours per machine = 300 × 16 × 0.85 = 4,080 h
N = 4,687.5 / 4,080 = 1.15 → 2 machines
Efficiency of the two machines = 4,687.5 / (2 × 4,080) = 57.4% of effective capacity; utilisation = 4,687.5 / 9,600 = 48.8% of the rated hours — spare capacity for the second product or for growth.

Example 3 - Assembly line balancing. Tasks (seconds): A 25, B 30, C 20, D 35, E 15, F 40, G 25, H 30 (Σ = 220 s). Precedence: A first; B and D after A; G after B; E and F after D; C after G; H last. Required output 450 units in a shift of 450 productive minutes.

Cycle time = (450 × 60) / 450 = 60 s
Nmin = ⌈220/60⌉ = ⌈3.67⌉ = 4 stations
Assignment (largest candidate, precedence respected): S1 = A + D = 60 s; S2 = B + G = 55 s; S3 = F + E = 55 s; S4 = C + H = 50 s (total 220 s ✓)
Line efficiency = 220 / (4 × 60) = 91.7%; balance delay = 8.3%
Smoothness index = √(0² + 5² + 5² + 10²) = √150 = 12.25 s
Output = 27,000 / 60 = 450 units/shift; station S1 is the pacing station — any delay there cuts output directly.

Example 4 - Johnson's rule (5 jobs, turning then grinding, hours).

Job	M1 turning	M2 grinding
A	5	2
B	1	6
C	9	7
D	3	8
E	10	4

Smallest time 1 (B, M1) → first; next 2 (A, M2) → last; next 3 (D, M1) → second; next 4 (E, M2) → second-last; C in the middle. Sequence B-D-C-E-A.
M1: B 0-1, D 1-4, C 4-13, E 13-23, A 23-28. M2: B 1-7, D 7-15, C 15-22, E 23-27, A 28-30.
Makespan = 30 h; M2 busy 27 h, idle 3 h (0-1, 22-23, 27-28); M1 idle 2 h at the end.
Compare FCFS A-B-C-D-E: M1 ends at 28 h, M2 runs 5-7, 7-13, 15-22, 22-30, 30-34 → makespan 34 h. Johnson saves 4 h (11.8%) with no extra resource.

Example 5 - SPT versus EDD on one machine (days).

Job	Processing time	Due date
1	4	6
2	2	4
3	6	18
4	3	8
5	5	12

SPT order 2-4-1-5-3: completions 2, 5, 9, 14, 20 → ΣC = 50, mean flow time 10.0 days; tardiness 0, 0, 3, 2, 2 → total 7, maximum tardiness 3 days.
EDD order 2-1-4-5-3: completions 2, 6, 9, 14, 20 → ΣC = 51, mean flow time 10.2 days; tardiness 0, 0, 1, 2, 2 → total 5, maximum tardiness 2 days.
Confirms the standard results: SPT gives the lowest mean flow time (and lowest average WIP by Little's law); EDD gives the lowest maximum tardiness — choose by what the customer penalises.
Dynamic check on day 0 for job 5: CR = (12 − 0)/5 = 2.4 (ahead of schedule); job 1: CR = 6/4 = 1.5; job 2: CR = 4/2 = 2.0 → job 1 is most at risk.

Aspect	Job production	Batch production	Mass / flow production
Layout	Process	Process or cell	Product (line)
PPC difficulty	Highest (every order unique)	Medium (lot sizing + sequencing)	Low after balancing
Key PPC tool	Gantt chart, priority rules	EBQ, MPS, CRP	Line balancing, takt, kanban feed
WIP and lead time	High, variable	Moderate	Low, predictable
Indian example	L&T heavy fabrication	Pump and valve makers	Maruti Manesar, TVS Hosur

Where it is used

Automobile and two-wheeler assembly: takt-based line balancing, mixed-model sequencing and part-feeding windows at Maruti Suzuki, Tata Motors, Hero and TVS; supplier schedules released from the MPS.
Capital goods and heavy engineering: BHEL, HAL, L&T — project-type production, backward scheduling from contractual delivery dates, bottleneck management on large machine tools.
Process industry: steel (SAIL, Tata Steel), cement and refineries — campaign planning, grade sequencing to cut changeover, and planned-shutdown scheduling.
Auto-component and SME manufacturing (Pune, Chennai, Coimbatore, Ludhiana clusters): batch sizing against setup time, finite loading on a few CNCs, and visual Gantt boards.
Job roles: PPC engineer/executive, production planner, master scheduler, materials planner, demand planner, SAP PP/APO functional consultant, operations and supply-chain analyst; a common entry role for mechanical graduates.
Beyond factories: hospital OT scheduling, airline crew and maintenance slotting, construction and MRO workshops use the same loading and sequencing logic.

Interview questions

Q: What is the difference between routing and scheduling? A: Routing fixes the sequence of operations and the machines/work-centres used; scheduling fixes the calendar time at which each of those operations starts and finishes.
Q: Which priority rule would you use to reduce average job delay in a loaded shop, and why? A: SPT, because clearing short jobs first minimises mean flow time and mean lateness — but add an ageing or slack override so long jobs do not starve.
Q: When does EDD beat SPT? A: When the penalty is on the worst-late job rather than on average delay — EDD minimises maximum tardiness.
Q: State Johnson's rule and its limitation. A: For n jobs on two machines processed in the same order, repeatedly take the smallest remaining time: on machine 1 schedule that job earliest, on machine 2 latest. It is optimal only for 2 machines (3 only under a special condition) and assumes no job passing, no breakdowns and known times.
Q: How do you decide the cycle time of an assembly line? A: Divide the available productive time per shift by the units required per shift; it must be at least the longest single task time, otherwise that task has to be split or duplicated.
Q: Line efficiency is 80% on a 5-station line. What does that tell you? A: 20% balance delay — one-fifth of the paid station time is idle; rebalance tasks, split the pacing station or re-set the cycle time.
Q: Difference between aggregate planning and the master production schedule? A: The aggregate plan sets total output, workforce and inventory in aggregate units over months; the MPS breaks that into specific end items per week and is capacity-checked before release.
Q: What is the difference between lateness and tardiness? A: Lateness is completion minus due date and can be negative for an early job; tardiness is the positive part only, so early jobs score zero.
Q: Forward or backward scheduling for a customer order with a firm delivery date? A: Backward from the due date to minimise inventory and finish just in time, but add a buffer at the bottleneck because backward schedules carry no slack.
Q: Your plant's throughput is below plan although average utilisation is only 70%. What do you check first? A: The bottleneck — average utilisation hides a saturated resource; check setup losses, breakdowns, scrap and starvation at the constraint, then schedule that resource first.
Q: How does batch size affect lead time? A: By Little's law, bigger batches mean more WIP waiting in queues, so flow time rises roughly in proportion; smaller batches need setup reduction to stay economical.
Q: How would you plan an extra 20% output for one month? A: Quantify the gap in machine-hours at the bottleneck, then rank options by cost — overtime (double wages by statute, worker consent required), a third shift, subcontracting, de-bottlenecking a single operation — and recheck material and manpower availability before promising it.

Common mistakes

Rounding the theoretical minimum number of stations down — it must always be rounded up.
Using the *average* station time instead of the longest station time to compute line output.
Confusing utilisation with efficiency, or quoting capacity without saying whether it is rated or effective.
Forgetting scrap/yield when converting required output into input quantity and machine-hours.
Treating tardiness as able to be negative, or averaging lateness and calling it tardiness.
Applying Johnson's rule to three or more machines without checking its validity condition, or to a job shop with different machine orders.
In exams, giving the sequence but not the makespan, idle times and Gantt chart that carry most of the marks; in practice, planning capacity on paper while ignoring maintenance downtime, setup time and statutory shift limits.