15.4 TPM, SMED, Standardized Work, and Visual Management

A lean programme is only real when a shop-floor tool changes a number the planning system uses. ECM Section IX.B names the tools; the exam tests what each does to a planning parameter. Four carry most of the weight: total productive maintenance, quick changeover, standardized work, and work-area design.

Total Productive Maintenance

Total productive maintenance (TPM) treats maintenance as an operator-owned reliability programme, not a repair department you telephone after a breakdown. Two pillars carry the exam items:

• Autonomous maintenance - operators own cleaning, lubrication, inspection, and fastener checks on their own equipment. Cleaning is inspection: the operator who wipes the machine down finds the leak while it is still a leak.
• Planned (preventive) maintenance - service intervals driven by time or usage, scheduled into the capacity plan as a known load instead of arriving as random downtime.

The planning consequence is blunt. An unplanned breakdown on a constraint is throughput you never recover: an hour lost on a non-constraint is absorbed by protective capacity, but an hour lost on the constraint is an hour lost by the whole plant. TPM is therefore a capacity tool, and preventive maintenance windows belong on the constraint's available-capacity calendar before work is loaded to it.

Overall equipment effectiveness

Overall equipment effectiveness (OEE) = Availability x Performance x Quality rolls the three loss families - downtime, speed, and defects - into one percentage.

A work center has 480 minutes of planned production time and loses 60 minutes to a die failure and a material stockout. It runs 380 units in the time left, against an ideal cycle time of 1.0 minute per unit, and 370 pass inspection.

Cross-check it: in 480 minutes at a 1.0-minute cycle the resource could have made 480 good units and made 370, and 370 / 480 = 77.1%. OEE is good output as a fraction of theoretical output.

Now the exam point. A 77% OEE means that eight-hour resource really delivers about 6.2 hours of good output (8 x 0.771), so loading it with eight standard hours overloads it by roughly 30% before the first schedule change. That is why available capacity must be derated by utilization and efficiency factors to reach rated capacity before you load it in capacity requirements planning; OEE is where the evidence for those factors comes from. (Where real output history exists, load against demonstrated capacity instead — that figure is measured from past earned hours, not calculated by derating.) The three factors multiply: averaging them gives a comfortable and wrong 92%.

Single-Minute Exchange of Die

Single-minute exchange of die (SMED), also called quick changeover, drives setup time toward single digits of minutes. The distinction the exam tests is:

• Internal setup - work that can only be done while the machine is stopped: unbolting the old die, mounting and aligning the new one, first-article verification.
• External setup - work that can be done while the machine still runs the previous order: staging dies at the machine, pre-heating, pre-kitting fasteners, pre-setting tool holders offline.

The method runs in that order: separate internal from external, convert internal work to external, then streamline what internal work remains with quick clamps, standardized shut heights, and no trial-and-adjust cycles.

Worked before and after

A press changeover takes 90 minutes, and video study shows 60 of them are fetching dies, hunting tools, staging material, and pre-heating - all doable while the press still runs. Converting that work to external leaves 30 minutes internal; quick clamps and killing the adjustment trials remove another 10, giving a 20-minute changeover. Setup is valued at $150 per hour, annual demand is 24,000 units, carrying cost is $6 per unit per year.

Read the last two rows together: you more than doubled the number of runs and still cut annual setup hours nearly in half, while average cycle stock fell 53%. Shorter setups make smaller lots economic, and smaller lots shorten manufacturing lead time, release cash tied up in cycle stock, and make mixed-model and level schedules feasible at all. That is why SMED is a planning tool and not a maintenance trick: setup time feeds lot sizing, available capacity, and the setup element of the routing. Named trap - lot size moves with the square root of setup cost, so cutting S by a factor of 4.5 cuts EOQ by only about 2.1, never proportionally.

Standardized Work

Standardized work is the documented current best sequence for a task, built from three elements:

1. Takt time - the customer demand rate the work content is paced to.
2. Work sequence - the exact order of the operator's steps.
3. Standard work-in-process (SWIP) - the minimum in-process quantity needed to run that sequence without stopping.

The principle the exam wants: you cannot improve an unstandardized process, because there is no baseline to measure against. If three operators run three different sequences, a claimed 12% gain is indistinguishable from operator variation. Standardized work is not a straitjacket - it is the object kaizen improves, and each improvement becomes the new standard, which is also what makes the run times in your routings credible.

Designing the Workflow and Work Area

5S

• Sort - remove what current work does not need, so dead tooling and obsolete work-in-process stop consuming space and cycle-count time.
• Set in order - a marked location for everything, turning retrieval time from a variable into a constant inside your standard time.
• Shine - clean as inspection; this is where 5S meets autonomous maintenance and protects availability.
• Standardize - one written convention across every cell, so the layout of one machine predicts the next and cross-training gets cheaper.
• Sustain - audits and discipline that hold the gain. Sustain is where most programmes fail, because the first four S's are an event and the fifth is a habit.

Visual management, andon, and jidoka

Visual management makes the state of the process readable in seconds by anyone standing in the area: min and max lines, kanban squares, colour-coded status boards, an hourly plan-versus-actual chart. The test is whether a stranger can tell the area is behind without asking.

An andon is the escalation half: a light, board, or cord signals an abnormality and calls help at the moment of the problem, not at the end of the shift when the count comes up short. The value is response time, not the light.

Jidoka, or autonomation, is automation with a human touch: equipment built to detect an abnormal condition and stop itself. The defect is contained instead of repeating through the batch, and because the machine no longer needs watching, one operator can tend several. The machine stops, the andon calls, and the problem gets root-caused with the tools in section 15.3.

Poka-yoke (mistake-proofing) is covered in section 14.1, and the exam puts the two in one answer set: poka-yoke prevents or detects the error at the point of work, while jidoka is the machine stopping itself. Automation and layout selection (section 3.3) support flow rather than create it - automating a process you have not first standardized and mistake-proofed only produces defects faster.

What Each Tool Actually Moves

Every row ends in a number you already use: setup time, run rate, yield, available capacity, lead time. A lean gain is banked only when it is re-entered into the item master, the routing, and the resource calendar - a 20-minute changeover still shown as 90 minutes keeps generating 1,342-unit orders, and a 77% OEE work center loaded as though it delivers eight clean hours keeps missing dates. Lean results show up as changed MRP and capacity parameters, not as posters on a wall.