故事 · 一台「忙得团团转」却产不出活的机器
Origin Story · The Machine That Was Always Busy But Never Productive
1960 年代日本中嶋清一推动 TPM 时发现一个尴尬现象:经理们盯着「机器开机率」自我感觉良好,可月底产量就是不达标。
问题在于「开着」不等于「产出好活」。机器可能在停机等修、在降速空转、在造废品 —— 三种损失互相叠加。
于是 OEE 把三者拧成一个数:可用率 × 性能 × 质量。一旦相乘,损失就被放大了 —— 三个 90% 乘起来只有 73%。
这个残酷的乘法,逼着工厂直面六大损失,而不是用单一指标自我安慰。
In the 1960s, when Seiichi Nakajima was building TPM in Japan, he spotted an awkward pattern: managers stared at machine uptime, felt good about themselves, and still missed the monthly output target.
The problem was simple — running is not the same as producing good parts. A machine can sit idle waiting for repair, limp along below cycle speed, or churn out scrap — three losses stacking on top of each other.
So Nakajima rolled all three into a single number: Availability × Performance × Quality. The moment you multiply them, losses get amplified — three 90% rates only yield 73%.
That brutal multiplication forces a factory to face the Six Big Losses head-on, instead of comforting itself with a single flattering metric.
1 损失瀑布:100% 理论产能怎么被一层层吃掉 Loss Waterfall: How 100% Theoretical Capacity Is Eaten Layer by Layer
OEE —2 乘积放大短板:三个率连乘只会更小 The Product Amplifies the Weak Link
三个率单独看都还行,可乘起来的 OEE 一定低于其中最低的那个。哪个最低,先改哪个 —— 这就是为什么 OEE 能精准指向瓶颈。 Each rate looks decent on its own, but the product always sits below the lowest of the three. Whichever rate is lowest is the one to fix first — that's why OEE points so cleanly at the bottleneck.
3 现实里的 OEE OEE in the Real World
TPM 全员维护:OEE 是 TPM 的总分牌。自主保全把小停顿和故障降下来,可用率 A 先涨起来。
TPM (Total Productive Maintenance): OEE is TPM's scorecard. Autonomous maintenance knocks out minor stops and breakdowns first, so Availability (A) climbs before anything else.
世界级 85%:A 90% × P 95% × Q 99.9% ≈ 85%,是离散制造业公认的卓越线;多数工厂在 40~60%。
World-class 85%: A 90% × P 95% × Q 99.9% ≈ 85% — the recognized excellence line in discrete manufacturing. Most plants live in the 40–60% range.
六大损失:停机两类(故障、换型)、速度两类(小停顿、降速)、质量两类(废品、启动报废),逐一对应 A/P/Q。
The Six Big Losses: two downtime losses (breakdowns, changeovers), two speed losses (minor stops, reduced speed), two quality losses (scrap, start-up rejects) — each maps cleanly onto A / P / Q.
瓶颈设备优先:先盯瓶颈工序的 OEE,那里每提升 1% 都直接转化成全厂产能,非瓶颈提了也白提。
Focus on the bottleneck: Only the bottleneck station's OEE drives plant throughput. Every 1% gained there flows straight to the top line; gains anywhere else are wasted effort.
一句话In One Line
OEE 用乘法而不是平均,是刻意的。平均会掩盖短板(90、95、60 的平均是 82,看着还行),
可乘积 0.90 × 0.95 × 0.60 = 51% 立刻暴露真相 —— 最弱的那一环按比例侵蚀全局。
这逼着你不能靠某一项的高分去补另一项的烂分,必须三条战线同时守住。
看懂这个乘法,你就明白为什么世界级工厂死磕每一个百分点:在乘积里,1% 的损失,会被另外两个率成倍放大。
OEE uses multiplication, not an average, on purpose. An average hides weakness — the mean of 90, 95, 60 is a comfortable 82.
But the product 0.90 × 0.95 × 0.60 = 51% tells the truth immediately: the weakest rate eats into the whole, proportionally.
That forces you to defend all three fronts at once — you cannot paper over one weak rate with a strong one elsewhere.
Once you internalize this multiplication, you understand why world-class plants fight for every percentage point: inside a product, a 1% loss gets amplified by the other two rates.
常见误用Common Mistakes
把三个率相加 / 取平均。OEE 是连乘,不是平均;平均会粉饰短板,乘积才暴露真实损失。
Adding or averaging the three rates. OEE is a product, not a mean. Averages flatter the weak link; the product exposes the real loss.
盲目追求全厂 OEE 最高。只有瓶颈设备的 OEE 决定产能,非瓶颈刷高分是过度生产(浪费)。
Chasing the highest possible OEE everywhere. Only the bottleneck's OEE governs plant throughput. Pushing non-bottlenecks higher just creates overproduction — a waste.
用 OEE 横向比不同设备 / 工厂。计算口径(标准节拍、计划停机算不算)各家不同,OEE 适合自己跟自己比趋势。
Benchmarking OEE across different machines or plants. Definitions (ideal cycle time, whether planned downtime counts) vary by site. OEE is a trend tool — compare yourself against yourself.