Question 3 of 7: Cells, Flexible Manufacturing Systems, Just-in-Time and Plant Networks

Question 3: Cells, Flexible Manufacturing Systems, Just-in-Time and Plant Networks (20 marks)

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

Part (a) — the manufacturing cell. A manufacturing cell is a group of dissimilar machines, laid out close together and dedicated to producing a family of parts that share similar processing requirements, with the work flowing from machine to machine within the group until the part is complete. The organising idea behind it is group technology: parts are classified by shape, size, material and required operations, and the classification exposes families whose members can all be made on the same small set of machines. The cell is then laid out in the sequence those parts require, commonly in a U shape so that one operator can tend several machines and so that the load and unload points are adjacent.

Cells were developed to escape the penalties of the traditional process layout, in which all lathes stand in one department, all mills in another, and so on. That layout maximises machine utilisation and supervisory specialisation, but a part must queue at every department, so the ratio of throughput time to actual cutting time is commonly fifty to one or worse. It also generates enormous material handling, large work-in-process inventory, long feedback loops when a defect appears, and scheduling complexity because each part has a different path. A cell restores the flow of a production line without requiring the volume of a production line: because the machines are adjacent and dedicated, transfer batches shrink towards one piece, throughput time falls from weeks to hours, work-in-process falls with it, setup times drop because tooling is shared across a similar family, and a defect is detected at the next operation rather than a week later. Quality and schedule accountability also become local, since the cell team owns a complete part.

Part (b) — the principle of the flexible manufacturing system. A flexible manufacturing system (FMS) is an automated cell: a group of computer-numerically-controlled machine tools or workstations, joined by an automated material-handling system (rail-guided or automatically guided vehicles, or a conveyor and pallet system), all under the supervision of a central computer that schedules the work, routes each pallet, downloads the part program and monitors tool life. The governing principle is that flexibility is bought by putting the changeover in software rather than in hardware. Each machine carries a large tool magazine holding the tools required by the whole part family, and each part is mounted on a standard pallet with a fixture that the handling system can move to any machine. Changing from one part to another therefore requires only that the supervisory computer call a different program and a different tool set — no physical setup, and hence essentially no changeover time. Routing flexibility follows from the same arrangement: if two machines can both perform an operation, the computer sends the pallet to whichever is free, so the system degrades gracefully when a machine goes down.

The capital cost is large for reasons that follow directly from that principle. The machines themselves are multi-axis machining centres with automatic tool changers and large magazines, each costing several times an equivalent stand-alone machine. Duplicate tooling must be provided so that alternative routes really exist, and pallets and fixtures — which must be accurate, rigid and identical — are required in numbers, not singly. The automated handling system, the load and unload stations, the wash and inspection stations, the central and cell controllers, and the network that ties them together are all additional. On top of the hardware sits a substantial engineering cost: part programs must be written and proven for every part in the family, fixtures designed, the scheduling logic configured and the whole system commissioned and debugged, which typically takes many months during which the investment earns nothing. An FMS is therefore justified only where the mid-volume, mid-variety region of the product spectrum is large enough to keep it loaded across several shifts.

Part (c) — why an FMS handles a wide range of lot sizes. The economic lot size exists only because a changeover consumes time that must be amortised over the pieces made after it; the classical batch-size formula is driven by the ratio of setup cost to carrying cost. In an FMS the changeover is a program call and a tool selection, so the setup cost tends to zero, and with it the economic incentive to batch. The system can therefore run a lot of one piece without penalty, and can interleave lots of one, ten and a thousand pieces in the same day, because the pallets carry their own fixtures and the controller knows which program belongs to which pallet. Two further features reinforce this. The tool magazines hold the tooling for the whole family simultaneously, so no physical tool change is needed between parts, and the automated handling system routes each pallet independently, so parts need not be grouped to travel together. The practical consequence is that an FMS occupies the middle ground of the volume–variety spectrum — below the volume at which a dedicated transfer line is cheaper, above the variety at which a manual job shop is cheaper — and covers a range of lot sizes that neither of those alternatives can span.

Part (d) — just-in-time and the pull principle. Just-in-time production aims to produce and deliver each item exactly when it is needed, in exactly the quantity needed, and its benefits follow from the inventory it removes. Work-in-process and finished-goods inventory fall sharply, releasing working capital and floor space and cutting the carrying, handling, obsolescence and damage costs that inventory attracts. Throughput time falls with the queues, which improves responsiveness to changes in customer demand and shortens the cash cycle. Quality improves because a defect is discovered at the next operation, within minutes, rather than after a buffer of a thousand parts has been produced — small lots make the feedback loop short and the containment cheap. The small-lot discipline also forces setup-time reduction, which in turn makes small lots economic, and the reduced buffers expose the problems — unreliable machines, unreliable suppliers, unbalanced lines — that inventory previously concealed, so the system drives continuous improvement rather than merely recording it. Space, handling equipment and scheduling paperwork all shrink.

It is called a pull system because authority to produce originates at the downstream end. In a conventional push system a central schedule computes what each work centre should make from a forecast and a master production schedule, and each centre produces to that plan and pushes the output forward whether or not the next centre needs it; a forecast error therefore becomes inventory. Under just-in-time, no work centre produces anything until the following centre signals consumption, classically by returning a kanban card or an empty container. The signal travels upstream from final assembly to sub-assembly to fabrication to the supplier, so the actual rate of consumption — not a forecast — sets the rate of production at every stage. Output is thus drawn, or pulled, through the plant by real demand, and inventory cannot accumulate anywhere, because the number of cards in circulation places a hard ceiling on the work-in-process between any two stages.

Part (e) — the function of a local area network. A local area network is a high-speed data communication system that interconnects the computing devices within a single plant or site — cell controllers, CNC machine tools, programmable logic controllers, coordinate measuring machines, automated storage and retrieval controllers, engineering workstations and the host production-control computer — using a shared physical medium and a common protocol under the owner's control. Its function in an integrated manufacturing plant is to make the plant's data a single shared resource rather than a set of islands. Concretely, it distributes part programs from the engineering database to the machine that needs them, so a revision made once is current everywhere; it carries production counts, machine status, tool-life data, downtime reasons and inspection results upstream to the host, so that scheduling and quality systems act on live data; it carries schedules, work orders and priority changes downstream to the cells; and it lets expensive peripherals and databases be shared instead of duplicated. Because manufacturing traffic is a mixture of short, time-critical status messages and large program transfers, plant networks are engineered for deterministic access and for electrical noise immunity, which is why token-passing and, latterly, switched industrial Ethernet with real-time extensions displaced contention-based schemes on the factory floor. The local area network is, in short, the nervous system that makes computer-integrated manufacturing integrated at all: without it the individual automated islands cannot co-ordinate, and the plant reverts to paper.