The procedure follows the logic of the standard method and adds what assembly of a geared unit requires. Figure 1 shows its six stages as a cycle, because a future state, once reached, becomes the current state for the next round.
The six stages of the proposed procedure, shown as a repeating improvement cycle
A line usually builds several variants that differ in ratio, housing or input arrangement. The map should follow one family, defined as the variants that pass through the same sequence of stations with similar work content. A product and process matrix, with variants as rows and stations as columns, makes the grouping visible. The scope should run from the point at which kitted or line-side components are presented to the first station, to the dispatch area after final test. Machining of gears and housings belongs to a separate map, although its delivery pattern enters this one as the supplier. A small team should be named: the line supervisor, an operator, a quality engineer and someone from planning, led by a person with authority to make changes.
The current state is drawn by walking the line, starting at dispatch and working upstream so that the map follows the pull of the customer. Each station or group of stations with continuous flow becomes a process box. Between boxes, the units waiting are counted and drawn as inventory. Table 1 lists the data to be recorded in each box and how each should be obtained. The rule that matters most is that figures are taken on the floor, with a stopwatch and a count, and not copied from routing sheets or the planning system, whose standard times describe an intended method that may no longer be the one in use (Rother and Shook, 1999).
Data to be recorded for each process box when mapping a gearbox assembly line
| Data item | Definition | How to obtain it |
|---|---|---|
| Cycle time | Interval between successive completed units at the station | Time ten or more consecutive cycles; note the range |
| Operator work content | Manual time per unit, separated from machine time | Time study by work element |
| Changeover time | Last good unit of one variant to first good unit of the next | Observe a real variant change |
| Uptime | Share of scheduled time the station can run | Stoppage log kept over several shifts |
| First-pass yield | Share of units passing without rework or adjustment | Count at the station and at final test |
| Inventory | Units waiting before the station, including rework | Physical count during the walk |
| Information trigger | What tells the station which variant to build next | Ask the operator; follow the paper or signal |
Timing should follow established work measurement practice: the job is divided into elements, several cycles are timed, and delays are recorded separately so that they are not hidden in an average (Freivalds and Niebel, 2013). Gearbox assembly needs three additions to the standard box. Where shims are selected by measurement, the number of trial fits per unit should be recorded, since a second or third fit multiplies the station time. The rework loop from the end-of-line test should be drawn as a flow of its own, with its inventory and its time. And the sub-assembly branches for shafts and differential or planetary sets should each be mapped back to their start, following the approach of Braglia et al. (2006), so that the branch that paces the main line can be identified.
The map is completed with the information flow, showing how customer orders become a build sequence and how each station learns of it, and with a timeline along the foot of the sheet that sets the sum of the processing times against the total lead time. Inventory is converted to time by dividing each count by the daily demand.
Takt time is calculated from the demand for the family and the working time available after planned breaks. Each station's cycle time is then set beside it on an operator balance chart. Stations above takt are the immediate constraints. Stations far below it hold spare capacity that may be combined. The search for the true constraint should take account of uptime and yield, since a station with an acceptable cycle time and frequent stoppages may limit output more than a slow and steady one. This is the reasoning of constraint management (Goldratt and Cox, 1984), and Hopp and Spearman (2008) show formally how variability in process times and outages lengthens queues and cycle times even when average capacity appears sufficient.
The analysis then asks, for each element of work, whether it adds value, is necessary without adding value, or is waste in the sense of Ohno (1988). In assembly the common findings are walking to fetch parts, searching for the right shim or seal, waiting for a press or a test stand, handling units into and out of buffers, and repeating work after a failed test. Each should be marked on the map where it occurs, with its time.
The future state is designed by answering in order the questions that Rother and Shook (1999) pose: what the takt time is, where continuous flow is possible, where supermarkets and pull signals are needed because it is not, which single process should be scheduled as the pacemaker, how the variant mix will be levelled there, and what improvements to individual processes the design presupposes. For a gearbox line the pacemaker is normally the main assembly line, with sub-assembly branches replenishing small supermarkets by pull.
Within the main line, work elements are redistributed so that each operator's content is close to takt and no station exceeds it. This is the assembly line balancing problem, for which a large body of exact and heuristic methods exists (Scholl and Becker, 2006; Boysen et al., 2007); for a manual line of modest length a balance chart and the guidance of Rother and Harris (2001) on cell design are usually sufficient. Process improvements should be aimed at the causes found in Stage 3. Typical candidates are kitting or sequenced line-side presentation to remove walking and searching, error-proofing devices that prevent a wrong bearing or a missing circlip from reaching the test stand (Shingo, 1986), quicker changeover of press tooling between variants, and gauging that gives the correct shim at the first fit.
A future state map is a design and not yet a plan. It should be divided into loops, each a segment of the stream that can be changed and stabilised on its own: the pacemaker loop around main assembly and test, a loop for each sub-assembly branch, and a supplier loop for incoming components. For each loop the team sets a measurable target, a date and a responsible person. Work normally starts at the pacemaker, because improvements there determine what the upstream loops must supply.
After each loop is changed, the data of Table 1 are taken again by the same method, and the new cycle time is compared with takt over a period long enough to include variant changes and ordinary disturbances. Where the target is met, the method is written into standardised work: the sequence of elements, the takt time and the standard in-process stock (Monden, 1998). Where it is not, the team returns to the analysis. The revised map then becomes the current state for the next cycle.