起源故事 · 中嶋清一与全员生产维护
Origin Story · Seiichi Nakajima and Total Productive Maintenance
1971 年,日本工程师 中嶋清一(Seiichi Nakajima)在日本电装(Denso)总结出 TPM(全员生产维护):
维修不该只是维修工的事,操作设备的人最了解设备的脾气,应该让他们参与日常点检、清扫、保养,把故障消灭在萌芽。
而要衡量设备到底好不好用,他提出了 OEE(设备综合效率)——把产能损失归为六大类,
用一个数字回答"这台机器真正在创造价值的时间,占了多少"。从此"设备明明开着却不出活"这件事,第一次被量化、被攻克。
In 1971, Japanese engineer Seiichi Nakajima formalised TPM (Total Productive Maintenance) while working with Denso:
maintenance shouldn't be left to the maintenance crew alone — the people who run the equipment understand its quirks better than anyone, so they should own the daily checks, cleaning, and lubrication and kill failures before they grow into breakdowns.
To measure whether a machine was actually pulling its weight, he proposed OEE (Overall Equipment Effectiveness) — rolling all capacity losses into six big buckets and answering, in a single number, how much of the time a machine is genuinely creating value.
From that moment on, the maddening problem of "the equipment is running but nothing's coming out" finally had a number on it, and a way to attack it.
1 产能瀑布:从 100% 一级级跌到实际 OEE Capacity Waterfall: From 100% Down to Actual OEE
六大损失six big losses2 三连乘:可用率 × 性能 × 质量 Triple Product: Availability × Performance × Quality
OEE 是三个百分比相乘,不是相加 —— 这就是为什么三项都还不错(比如各 90%)时,乘起来也只有 73%。任何一环掉链子,整条都被拖下水。 OEE is three percentages multiplied, not added — that's why even when all three look respectable (say, 90% each), the product is only 73%. One weak link drags the whole chain down.
3 现实里的 OEE 与 TPM OEE & TPM in the Real World
设备诊断:不知道该改哪台机器时,先测 OEE,最低的那台往往藏着最大的隐性损失。
Equipment diagnosis: not sure which machine to attack first? Measure OEE — the lowest scorer usually hides the fattest invisible losses.
自主保养:让操作工每天点检、清扫、紧固,TPM 把故障停机这一项压到最低。
Autonomous maintenance: hand daily inspection, cleaning, and tightening back to the operators and TPM drives the breakdown loss toward zero.
瓶颈提升:约束设备每提升 1% OEE,等于整厂多出 1% 产能,回报远超买新机器。
Constraint lift: every 1% of OEE you add at the bottleneck machine is 1% more capacity for the whole plant — the payback dwarfs buying another machine.
世界级标杆:OEE ≥ 85%(可用率 90% × 性能 95% × 质量 99%)被视为世界级水平。
World-class benchmark: OEE ≥ 85% (Availability 90% × Performance 95% × Quality 99%) is the line that separates world-class manufacturers from the rest.
一句话In One Line
"设备在转"不等于"设备在创造价值"。OEE 的威力,是把模糊的"它好像不太行"翻译成一个能被攻击的数字,
再用六大损失告诉你该攻哪里。它和换模(SMED)、防错(Poka-Yoke)、自働化(下一站的安灯)天然咬合 ——
换模快了,可用率涨;防错好了,质量涨。但记住:OEE 是诊断仪,不是 KPI 鞭子。
为了刷高数字而拼命赶产能、压保养,反而会埋下更大的故障。
"The machine is running" is not the same as "the machine is creating value." OEE's real power is turning a vague "it doesn't feel right" into a number you can attack,
and using the six losses to tell you exactly where to attack. It locks naturally into SMED (faster changeovers), Poka-Yoke (mistake-proofing), and jidoka (the andon station coming up next) —
faster changeovers lift availability; better mistake-proofing lifts quality. But remember: OEE is a diagnostic instrument, not a KPI whip.
Chasing the number by pushing production harder and starving maintenance just plants bigger breakdowns down the road.
常见误用Common Mistakes
把 OEE 当考核指标去追高。它是发现损失的诊断工具,追数字会逼出过度生产与少保养。
Turning OEE into a performance target to chase. It's a diagnostic tool for finding losses — chasing the number pushes operators into overproduction and skipped maintenance.
给非瓶颈设备也死磕 OEE。非约束机器提 OEE 只会多堆库存,先盯瓶颈设备。
Squeezing OEE on non-bottleneck equipment. Lifting OEE on a non-constraint machine just piles up inventory — work the constraint first.
三率分不清就乱报数。可用率管停机、性能管快慢、质量管好坏,错配会找错改善方向。
Mislabelling losses across the three rates. Availability handles downtime, Performance handles speed, Quality handles defects — bucket a loss wrong and you'll attack the wrong cause.