起源故事 · 从二战电子管到浴盆
Origin Story · From WWII Vacuum Tubes to the Bathtub
二战期间,雷达和飞机上的电子管频繁失效,军方第一次被迫系统研究"东西到底什么时候坏"。工程师们发现:
失效率不是匀速的 —— 新装的器件因制造缺陷早早夭折,熬过去后进入一段平稳的偶发故障期,到寿命末期又因老化磨损加速失效。
把失效率对时间画出来,恰似一个浴盆。MTBF 概括了平稳期的平均水平,但只有看懂整条浴盆曲线,
才能回答"保修期定多久、备件囤多少、什么时候该大修"。它和上节的威布尔分布说的是同一件事:β<1、=1、>1 正对应浴盆三段。
During WWII, vacuum tubes in radars and aircraft kept failing, and the military was forced — for the first time — to study systematically "when does stuff actually break?" The answer:
the hazard rate isn't uniform. Brand-new units die early because of latent manufacturing defects; survivors settle into a flat, random-failure plateau; near end of life, aging drives the rate back up.
Plotted against time, the curve looks like a bathtub. MTBF only summarizes the flat middle — but only the full bathtub answers
"how long do I warranty it, how many spares do I stock, when do I overhaul?" It's the same story as the Weibull distribution from the previous page: β<1, β=1, β>1 map directly onto the three stages.
1 浴盆曲线:设备现在站在一生的哪一段? The Bathtub Curve: Where in Its Life Is Your Equipment Right Now?
随机期Useful life蓝色是失效率 h(t):左边早夭递减、中间随机平坦、右边磨损递增。黄色标记是设备此刻的年龄 —— 它落在哪一段,就决定你该用什么维护策略。 Blue is the hazard rate h(t): decreasing on the left (infant mortality), flat in the middle (useful life), rising on the right (wear-out). The yellow marker is the equipment's current age — whichever zone it lands in dictates your maintenance play.
2 可靠度 R(t):还能撑多久不坏 Reliability R(t): How Much Longer Will It Survive?
随机期里可靠度按 R(t)=e⁻λᵗ 平滑衰减;一旦进入磨损期,失效率陡升,可靠度断崖式下跌 —— 这正是"到点就得换"的数学原因。 In useful life, reliability decays smoothly as R(t) = e⁻λᵗ. Once the part enters wear-out, the hazard rate spikes and reliability falls off a cliff — the math behind "replace it on schedule, not when it breaks."
3 现实里的浴盆曲线 The Bathtub Curve in the Field
保修期设计:把保修期定在"早夭期之后、磨损期之前",让缺陷品在售前老化筛选中暴露。
Warranty design: Set the warranty window to land "after infant mortality, before wear-out" so latent defects surface during pre-sale burn-in, not in the customer's hands.
预防性维护:磨损期失效率陡升前定期更换(轴承、刀具、轮胎),别等坏了才修。
Preventive maintenance: Replace bearings, tooling, and tires on a schedule — before the wear-out curve steepens. Don't wait for breakdown.
备件库存:随机期失效率恒定,可按 λ 算出稳态备件消耗,避免囤太多或断货。
Spares inventory: With a constant useful-life hazard, λ gives you the steady-state spares burn rate — no overstock, no stockouts.
老化筛选:出厂前给电子产品做 burn-in,主动熬过早夭期,交到客户手上时已进入平稳段。
Burn-in screening: Run electronics on a test bench before shipment to power through infant mortality in-house — the customer only ever sees the flat middle.
一句话In One Line
MTBF 只是浴盆平底那一段的平均失效率倒数,它假设"失效率恒定、不会老化"—— 这只在随机期成立。
拿 MTBF 去推磨损期设备的寿命,会严重高估可靠性,等于无视浴盆右壁的陡升。
真正的维护决策要看你现在站在浴盆哪一段:早夭期靠筛选、随机期靠备件、磨损期靠定期更换。
下一节我们把多个这样的部件组装成系统,看串联和并联如何改写整体可靠度。
MTBF is just the reciprocal of the average hazard rate across the flat floor of the tub. It assumes "hazard rate is constant, no aging" — true only in useful life.
Use MTBF to project the lifetime of wear-out equipment and you will badly overstate reliability — you are pretending the right-hand cliff doesn't exist.
Real maintenance decisions follow your position on the tub: infant mortality wants screening, useful life wants spares, wear-out wants scheduled replacement.
Next page: we wire several of these parts together into a system and see how series vs. parallel structures rewrite overall reliability.
常见误用Common Mistakes
把 MTBF 当"保证用到的寿命"。MTBF 是平均间隔,不是承诺寿命,半数会更早坏。
Treating MTBF as a guaranteed lifespan. MTBF is an average between failures, not a promise — roughly half the population will fail sooner.
对磨损期设备用恒定失效率算可靠度。磨损期 h(t) 递增,必须用威布尔 β>1。
Using a constant hazard rate to model wear-out parts. In wear-out, h(t) rises — you must switch to a Weibull model with β>1.
忽视早夭期直接发货。做老化筛选/burn-in,把婴儿死亡挡在出厂前。
Shipping units without screening. Run burn-in so infant mortality is caught in the factory, not the field.