Flexible manufacturing systems

Why it matters

Transfer lines are efficient but make one part; stand-alone CNC machines are flexible but need people to load them and move parts. A flexible manufacturing system (FMS) sits between the two: a group of CNC workstations, linked by automated material handling and run by a central computer, that can machine a mix of different parts in random order with almost no changeover time. Engine-block, gearbox and aerospace component plants use FMS for mid-volume, mid-variety work, and the simple bottleneck model lets you size one on paper.

Key ideas

Definition. An FMS is a highly automated group-technology cell: processing workstations (usually CNC machining centres), interconnected by an automated material handling and storage system, under integrated computer control, able to process different part styles simultaneously.

Tests of flexibility. A system is flexible if it can (1) process different part styles in a non-batch mode, (2) accept changes in the production schedule, (3) respond gracefully to machine breakdowns, and (4) accept new part designs. Systems that pass all four are true FMSs; if they need batches and changeovers they are only flexible in a limited sense.

Kinds of flexibility. Machine flexibility (variety of operations per machine), production (part-mix) flexibility, mix-change flexibility, product flexibility (new parts), routing flexibility (alternative machines), volume flexibility and expansion flexibility.

Types by size and by variety.

• By number of machines: single machine cell (one CNC machine with pallet storage), flexible manufacturing cell (two or three machines plus handling), flexible manufacturing system (four or more machines).
• By part variety: a dedicated FMS makes a limited, known set of parts, often with special-purpose machines; a random-order FMS handles a large and changing part family with general-purpose CNC machines and more sophisticated control.

Components.

• Workstations — CNC machining centres and turning centres, plus load/unload stations (where parts are fixtured on pallets), inspection (CMM), washing and assembly stations.
• Material handling and storage — primary handling between stations (AGVs, rail-guided vehicles, roller conveyors, robots), secondary handling at each machine (pallet changers, shuttles), and AS/RS or pallet stockers for buffering.
• Computer control — workstation control, distribution of NC programs (DNC), production control and dispatching, traffic control of vehicles, shuttle control, tool management (tool life, tool location), work-handling control, system monitoring and reporting.
• Human resources — loading/unloading, tool setting, maintenance, programming and management.

Layouts. In-line, loop, ladder, open-field, and robot-centred cell.

Benefits and limits. Higher machine utilisation than a job shop, less work-in-process, shorter lead times, lower direct labour, ability to run unattended through a third shift. Limits: very high capital cost, complex software, and limited true part variety — FMSs work best for a stable family of medium-volume parts.

The bottleneck model. For steady state, ignore queuing and assume each station is busy whenever it has work. The station whose workload per server is largest limits the production rate of the whole system; all other stations are under-used. Adding a server at the bottleneck raises output until another station becomes the bottleneck.

Formulas

WL_i = Σ_j p_j · t_ij

• WL_i workload of station i (min per part produced, averaged over the mix); p_j fraction of part j in the mix (Σ p_j = 1); t_ij processing time of part j at station i (min).

R_p* = s* / WL*

• R_p* maximum production rate of the system (parts/min); WL*/s* the largest value of WL_i/s_i over all stations (the bottleneck); s_i number of servers at station i.

R_pj* = p_j · R_p*

• Production rate of part j.

U_i = (WL_i / s_i) · R_p*

• Utilisation of station i (fraction); U = 1 at the bottleneck.

WL_t = n_t · t_t

• Workload of the transport system (min per part); n_t mean number of transports per part (stations visited minus 1, counting load/unload once at the start and once at the end); t_t mean transport time (min). Treat the vehicles as servers.

Worked examples

Example 1 (standard). One part type is processed: load/unload 6 min (1 server), milling 30 min (3 machining centres), inspection 8 min (1 CMM). Find the bottleneck, the maximum production rate and the utilisations.

1. WL/s: load/unload 6/1 = 6 min; milling 30/3 = 10 min; inspection 8/1 = 8 min.
2. Bottleneck = milling, 10 min per part. R_p* = s* / WL* = 1/10 part/min = 6 parts/h.
3. Utilisations: load/unload 6/10 = 60 %; milling 100 %; inspection 8/10 = 80 %.

Example 2 (GATE level). An FMS makes parts A (40 % of output) and B (60 %). Station 1 (load/unload, 1 server): 5 min for either part. Station 2 (milling, 2 servers): A 30 min, B 20 min. Station 3 (drilling, 1 server): A 12 min, B 15 min. Every part goes L/U → mill → drill → L/U, and 2 AGVs take 4 min per move. Find the system and part production rates and the utilisations.

1. WL_i = Σ_j p_j · t_ij: WL₁ = 5 min; WL₂ = 0.4 × 30 + 0.6 × 20 = 24 min; WL₃ = 0.4 × 12 + 0.6 × 15 = 13.8 min.
2. Per server: 5, 24/2 = 12, 13.8/1 = 13.8 min. Transport: n_t = 3 moves, WL_t = 3 × 4 = 12 min, per vehicle 6 min.
3. Bottleneck = drilling (13.8 min). R_p* = 1/13.8 = 0.0725 part/min = 4.35 parts/h.
4. R_pA = 0.4 × 4.35 = 1.74 parts/h; R_pB = 0.6 × 4.35 = 2.61 parts/h.
5. Utilisations: L/U 5/13.8 = 36.2 %; milling 12/13.8 = 87.0 %; drilling 100 %; AGVs 6/13.8 = 43.5 %.
6. Insight: a second drilling machine would move the bottleneck to milling (12 min) and raise output to 5 parts/h.

Common mistakes

• Adding up machine capacities instead of finding the bottleneck station.
• Forgetting to divide the workload by the number of servers.
• Using the part-mix fractions wrongly — they weight processing times, not machine counts.
• Forgetting the transport system can be the bottleneck.
• Treating the bottleneck rate as achievable output. Real systems lose output to queuing, breakdowns and starving, so the model gives an upper bound.

For GATE PI

Expect MCQs on FMS components, types (dedicated versus random-order), flexibility tests and layouts. Numericals use the bottleneck model: workloads with a part mix, the bottleneck station, maximum production rate, part production rates, utilisations, and the number of servers needed to reach a target rate. Practise organising the data in a station-by-part table.

Quick check

1. What are the four tests of flexibility?
2. Which FMS type suits a large, changing part family?
3. Stations: L/U 4 min (1 server), turning 25 min (2), milling 18 min (1). Find the maximum production rate.
4. In question 3, what is the utilisation of the turning station?
5. If a second milling machine is added in question 3, what is the new maximum rate?

Answers: 1. Part-variety, schedule-change, error-recovery and new-part tests. 2. Random-order FMS. 3. Bottleneck milling 18 min → 3.33 parts/h. 4. 12.5/18 = 69.4 %. 5. Bottleneck becomes turning (12.5 min) → 4.8 parts/h.