火箭这行有个残酷的规矩:及格线永远是 100%。
没有召回,没有补丁,没有灰度发布,也没有回滚。从点火那一刻起,要么 100%,要么归零。2015 年 7 月 20 日,SpaceX 就猎鹰 9 号(Falcon 9)CRS-7 任务失败召开电话会议,马斯克(Elon Musk)一上来就把这句话摆了出来。
此时距离那次爆炸过去不到一个月。猎鹰 9 号在给国际空间站送龙飞船(Dragon)的途中解体,残骸散落在佛罗里达外海。马斯克召集记者,通报初步调查结论。他反复强调,所有内容都是“初步评估”,后续调查可能完全推翻。
然后他开始拆解整个过程。
第一级火箭一切正常,问题出在第二级
起飞之后,第一级火箭工作正常,没有任何异常读数。第一级在整个上升段都按预期消耗推进剂、按预期关机分离,没有任何提前暴露的征兆。
故障发生在第二级。液氧箱里出现超压,结构失效,火箭解体。
这里有个容易被忽略的细节:第一级在第二级出事之后,还继续工作了好几秒,直到第二级解体导致结构断裂才失控。与此同时,龙飞船扛住了二级的超压,一直向地面发遥测信号,直到飞出海平面、超出地面站接收范围才失联。
从首次出现异常到全部数据丢失,前后 8.93 秒。
听起来很短,但调查团队花了好几个星期。一枚猎鹰 9 号上有超过 3000 个遥测通道,外加多路视频、物理碎片、地面雷达记录。要把所有这些数据对齐到毫秒级,再和外部视频逐帧匹配,本身就是件极难的事。马斯克在电话会上说,当毫秒都变得重要的时候,把所有东西严丝合缝地对上,难得超出想象。
一根支杆,在 3.2 个 G 的时候断了
故障点最后锁定在第二级液氧箱里的一根支杆(strut)。
这根支杆的作用,是固定住一个复合材料缠绕的氦气瓶。二级液氧箱里浸泡着好几个这样的高压氦气瓶,瓶内压力大约 5500 PSI。把它们泡在液氧里,是为了用低温把氦气冷却到接近液氧的温度,大幅提高氦气密度,让同样体积的瓶子装下更多氦气。
发动机点火之后,氦气从瓶里流出,经过发动机加热,再回到液氧箱和燃料箱,给贮箱增压。它的作用是补偿推进剂被消耗后腾出来的体积,同时维持结构所需的内部压力。换句话说,氦气系统一旦出事,整个二级就会失压、失稳。
初步判断是:飞行过程中,固定其中一个氦气瓶的支杆断裂。瓶子失去约束,在液氧箱里高速向上冲。
为什么会往上冲?因为火箭在加速,浮力随重力加速度(G 值)成正比增加。这个物理现象有点反直觉:多数人以为加速的时候箱子里的东西会往下沉,但在液体里浮力方向相反,加速度越大,密度低的物体被往上顶的力就越大。氦气瓶密度远低于液氧,G 值越高,它越想往上蹿。
事故发生时,二级承受大约 3.2 个 G。这根支杆就是在这个节点断的。
氦气瓶脱开之后,大量氦气泄进液氧箱。二级液氧箱里的气体体积本来只有 2% 左右,因为二级推进剂在入轨前基本不消耗,箱子里几乎全是液体。所以不需要漏多少氦气,压力就会迅速飙升,把液氧箱撑破。
这就是整条故障链。
设计扛一万磅,两千磅就断了
这根支杆的材料认证能力是承受 1 万磅的力。实际断裂时,大约只有 2000 磅。
五倍的差距。
更麻烦的是,这种支杆在猎鹰 9 号上已经飞了无数次。每枚火箭上都有好几百根,累计飞行历史有几千根。高清收尾照片里看不出任何装配错误或外观损伤。马斯克在电话会上说,正因为一切看起来都完美,才让这件事成了一个“谜”。
所以一开始,调查团队根本没有怀疑到它头上。
SpaceX 从供应商那里拿来一大批同款支杆做拉伸测试,头几批里最低的失效点也在 6000 磅。团队一度认为“不可能是支杆的问题”,转去别处找原因。找不到,数据解释不通。
后来又搞来一大批带同款螺栓的支杆,测试了很大的数量,终于找到一根在 2000 磅以下就断的。做了材料分析之后发现,这根支杆的钢材晶粒结构有问题,成型工艺不对,外观看起来完全正常,内部却已经不合格。
马斯克的原话是:这大概是某一批看起来好、实际上不好的螺栓流了进来。
地面测试为什么测不出来
有人问,这种问题地面测试怎么会发现不了?
答案跟前面说的浮力有关。
火箭在地面做静态点火(stage hold-down firing)的时候,G 值只有 1。支杆感受到的浮力也只有飞行时的几分之一。它在地面上完全正常。只有真正飞起来、加速度拉到 3 个 G 以上,浮力才会成倍放大,把支杆逼到极限。
要在地面完整复现这个受力环境,理论上得把整个二级放上一台巨大的离心机,转到飞行加速度。地球上没有这种设备。马斯克说得很直白:这种测试的终极形态在物理上不存在,你只能通过间接手段去逼近,而间接手段有时候不够用。
这次就是不够用。
这也是火箭工程最折磨人的地方。飞机的部件可以在地面台架上做疲劳测试,汽车的零件可以跑几十万公里验证。火箭不行。每次飞行都是一次性的、极端叠加的、不可复现的工况。你只能无限逼近它,不能真正覆盖它。
遥测里有一个先降后升的矛盾
调查过程中有一段数据让人费解。
氦气系统压力先是下降,这符合“管路破裂”的预期。但紧接着,压力又回升到接近初始值。如果氦气瓶真的破了,压力不该恢复。
团队目前的解释是:氦气瓶脱开之后在箱子里翻转、扭动,可能把连接瓶子和氦气歧管的管路给夹住了,于是氦气系统本身的压力恢复。但在这之前,已经有足够的氦气泄进液氧箱,把液氧箱撑爆。
这个解释马斯克自己也承认“有点推测成分”,是目前能想到的最合理版本。
另一个佐证来自声学三角定位。二级上分布着加速度计,通过高频事件到达不同传感器的时间差,可以反推断裂发生的具体位置。定位结果指向那根支撑支杆。
如果是氦气瓶内衬出现裂纹,氦气会持续泄漏,遥测上会看到更大量的氦气流失,而不是先降后升。这也支持“瓶子整体脱开、管路被夹住”的判断。
龙飞船本来可以活下来
讲龙飞船的时候,马斯克的语气沉了下来。
龙飞船扛住了二级超压,一直发遥测信号,直到飞出海平面、超出接收范围。它没有在爆炸中解体。
问题出在软件。
龙飞船 1 号是货运版,降落伞展开的触发软件根本没有装。如果当时软件能自动判断异常、开伞,龙飞船本可以落进海里,货物也能保住。
“这是整件事里最令人难过的部分。只差一点软件,龙飞船就回来了。”(马斯克)
SpaceX 的决定是,把原本给龙飞船 2 号(载人版)准备的应急软件提前装到龙飞船 1 号上。下一趟货运任务就会带上这个能力:一旦发射阶段出问题,龙飞船会自己尝试开伞自救。
有人问,载人版遇到这种情况会怎样?
龙飞船 2 号有发射逃逸推进器,系统会在异常发生的瞬间点火,把飞船推离火箭,等拉开安全距离之后再开伞落海。但即便没有逃逸推进器,龙飞船 1 号本身已经证明它能扛住二级爆炸。所以载人版的安全性只会更高。
马斯克用了一个词形容龙飞船:一个相当顽强的生物(a tough creature)。它在爆炸之后还能持续广播遥测,本身就说明舱体强度远超预期。
七年没失败,公司从 500 人变成 4000 人
有记者问,你之前说公司可能变得有点“自满”,具体指什么?
马斯克回答得很坦率。
猎鹰 1 号早期连续失败的那几年,整个团队被吓出了“极度偏执”。但那是七年前的事了。当时公司只有 500 人。现在有 4000 人。
也就是说,今天公司里绝大多数人,只见过成功。
“当你只见过成功,你就不会那么害怕失败。”(马斯克)
他提到一个习惯:每次发射前,他都会给全公司发邮件,说如果有人能想到任何推迟发射的理由,不管直属领导同不同意,立刻打他手机或发邮件。
这封邮件他每次都发。但发到第 20 次的时候,效果就变了。
“第 20 次发那封邮件的时候,大家会觉得,哦,马斯克又在偏执了。所以它可能不会再像以前那样有分量。”(马斯克)
这次失败之后,他说,全公司重新理解了把火箭送进轨道有多难。“我们会因此变得更强。”
供应商的螺栓,和一条新的质量规则
出问题的支杆不是 SpaceX 自己造的,来自外部供应商。马斯克拒绝公开供应商名字,理由是“只会带来互相指责”。
SpaceX 虽然自己造火箭的大多数主要部件,但外围零件有几百家供应商,全部自制的可能性不存在。
这次的教训催生了一条新规则:不再相信材料认证。
以前的做法是,供应商出具认证,证明这批支杆能扛 1 万磅,SpaceX 就按这个数采信。现在不行了。不管认证写的是什么,每一根支杆都要单独做拉拔测试(pull test)。
马斯克说,这会增加一些成本,但不认为会影响到火箭售价。他还提到,下一版设计会把螺栓材料换成铬镍铁合金(Inconel),具体还在评估。支杆本身不打算收回自制,会换一家供应商,换一种设计。
有人问要不要多加一根支杆做冗余,他的回答是不必要:一根支杆的设计载荷已经超过实际受力的两倍,再加一根没有意义,真正的问题在于单根质量不可靠,不在数量不够。
这条新规则的适用范围,也不只是支杆。马斯克说,过去 SpaceX 对主结构、主气瓶这类大件都会做独立测试,但对支撑类硬件、尤其是那些设计余量大、供应商认证看起来特别齐备的小件,往往是直接采信认证。从这次开始,所有支撑硬件都要做物理测试,“证书再好看也不信”。
九月之前不会飞,猎鹰重型推到明年春天
发射时间表是记者们最关心的问题。
马斯克的回答很谨慎:不早于九月。具体哪一次任务、哪个客户,都还没定,可能会有调整。
支杆本身的修复并不复杂:换更强的设计,逐根测试,理论上只会把发射推迟几个月。但 SpaceX 想先把所有数据翻一遍,确认诊断无误,排除其他潜在问题,再和 NASA、空军、FAA 以及商业客户对齐,才会公布确切日期。
商业载人时间表不受影响。马斯克说,商业载人的关键路径是龙飞船 2 号的设计和验证,跟这次故障没有直接关系。
猎鹰重型(Falcon Heavy)则被往后推了。原计划更早,现在看大概明年春天,四月左右。原因是资源集中在猎鹰 9 号上。
保险方面,SpaceX 不给自己的发射投保。马斯克说,火箭行业基本没人给发射本身投保,只有地面第三方损害这类极低概率项目才保。这次最大的损失是发射节奏被打断带来的收入递延,规模在“数亿美元”级别。
那一年,三次空间站补给失败
发布会快结束时,有人问到 NASA 即将授出的第二轮商业载人合同会不会受影响。
马斯克说这该问 NASA,但顺带提了一个背景:过去一年,国际空间站补给任务出了三次事故。SpaceX 一次,轨道科学公司(Orbital Sciences)一次,俄罗斯进步号(Progress)一次。三次的故障模式各不相同。
“火箭是极其困难的事。一般来说,当一个习语存在时,总有它的道理。”(马斯克)
这句话是整场发布会里最能概括主题的一句。火箭不是软件,不能热修复,不能灰度发布,不能回滚。每一次发射都是一次性的,所有零件同时工作,所有系统同时承压,任何一个微小的、在地面看不出来的缺陷,都可能在几百秒里把整枚火箭变成海面上的一片碎片。
而这一次,问题出在一根两英尺长、最粗处一英寸的钢制支杆上。它的认证强度是一万磅,实际断裂强度是两千磅。外观完美,内部晶粒结构不对。它在过去几百次飞行里都没出过问题,偏偏这一次,在 3.2 个 G 的时候断了。
这就是火箭。
给从业者的几条行动建议
- 对安全关键件,认证不等于可靠。只要条件允许,对哪怕长期稳定供货的小件也做单件实测,尤其是那些“设计余量极大、看起来永远不会出问题”的支撑类零件。事故往往就发生在你最有信心的地方。
- 警惕团队的“成功疲劳”。连续成功会降低警觉,新人只见过成功就不会害怕失败。可以定期把历史事故日志拿出来复盘,让团队保持对失败的真实体感。
- 软件兜底能力要前置,不要等下一代产品。Dragon 1 缺的那段开伞逻辑,原本计划给 Dragon 2,如果早一步装上去,这次损失会小得多。判断一个安全软件是不是可以提前移植,标准不是产品路线图,是“如果早一天上线,能救什么”。
- 故障复盘要把数据对齐到最小时间粒度。3000 个遥测通道加视频,全部压到毫秒级匹配,才有机会从“先降后升”这种反常信号里读出物理机制。模糊的时间对齐会直接抹掉真相。
- 善用声学/振动三角定位推断断裂位置。加速度计的高频事件时间差,能在没有实物残骸的情况下给出结构断裂的可信定位,是火箭、飞机这类一次性试验里性价比极高的诊断手段。
- 让“一票否决”通道真的有效。发射前给全员开放、可以绕过直属领导的直报通道很有价值,但同一个动作重复 20 次之后就会形式化。想办法给它换形式、换入口,让它不要变成“老板又在偏执”的例行公事。
- 供应商质量要穿透管理。外观和高清收尾照片都正常的零件,内部晶粒结构可能已经不合格。对高风险小件,建立进厂抽检加批次追溯机制,不要只依赖供应商的合格证书。
- 接受“无法在地面完全复现”的现实。离心机造不出来,就老老实实承认间接测试的局限,用更高的设计余量、逐件实测、飞行数据持续回归来补足,不要假装地面测试已经覆盖了所有工况。
0:01Good afternoon. My name is Sharon and I'll be your conference operator today. At this time, I would like to welcome everyone to the SpaceX update conference call. All lines have been placed on mute to prevent any background noise. After the speaker's remarks, there will be a question and answer session. If you'd like to ask a question during this time, simply press star, then the number one on your telephone keypad. If you'd like to withdraw your question, press a pound key. Thank you. Mr. Elon Musk, you may begin your conference.
0:24All right. Thank you. so at at the end of my bble comments we're going to post just a written version to the SpaceX website. So if you don't happen to catch something while you know during my bal discussion that you'll be able to look at that on in the written form immediately after the the comments end. okay so just going through this what the best of what we know thus far. and I do want to emphasize this is an initial assessment and further investigation may reveal may reveal more over time.
1:03so so I'll just run through it. prior to the the failure the first stage of the vehicle was operating normally. There was no no issues seen on the first stage roll. It actually continued to power through the over pressure event on the second stage for several seconds following the first stage over coming apart. In add
1:34in addition, Dragon actually survived not only the second stage overressor but continued to communicate until the vehicle dropped below the horizon and out of range. in fact if we had if the software had had the had initiated the parachute deployment then the Dragon spacecraft would be believed would have survived and for for future missions even for the cargo version of the Dragon spacecraft. We're now including contingency software that if something were to go wrong with the vehicle Dragon will will always attempt to save itself. this is this is certainly something that we have included in Dragon 2 which is the the sort of crew the next generation Dragon
2:16that will be carrying cargo and crew to the space station but we've now advanced that activity to include it in Dragon 1 as well. so that that's an unfortunate thing because we we really we could have saved Dragon if if we had the right software there. and so we we've we've been going through the investigation with daily participation and insight from the FAA, NASA, and the US Air Force. And I just like to say a word of appreciation for the help that we've received here. it's been great to have so much expertise reviewing the the flight data. and we're looking not just for issues that occurred that that caused this this failure, but any near misses that
3:00could potentially affect future flights. so we're looking very closely at saying just in general, is there anything that we can do to reduce the probability of failure with with future flights. the the fundamental nature of rocketry is is is that it is a case where a passing grade is 100% every time. when there's it's not possible to issue a recall or a patch
3:30or anything like that. from the moment of liftoff it's 100% or nothing. So so we're looking for any tiny thing that could improve things in the future. so the part of the reason why it's taken us a long time to at least come up with with where the preliminary assessment is that there are there's such an enormous amount of data. We've over 3,000 telemetry channels as well as multiple video channels and physical debris and the key events in this case happened very quickly from the first indication of an issue to loss of all data was just 893 seconds. so over
4:10the past few weeks we spent thousands of hours going through a painstaking process of matching up the data across all telemetry systems down to the millisecond because it's very important to understand the exact sequence of what happened and then matching that to external video and and and really
4:28when when milliseconds matter it's it's remarkably difficult to actually line everything up exa exactly right. and as as said the investigation remains ongoing and and we're continuing additional testing to fully validate these conclusions. so again this is just our preliminary preliminary analysis. and the so
4:51so the the preliminary conclusion is that the the failure arose from a a strut in the second stage liquid auction tank. there was a strut that was holding down one of the composite over wrapped helium pressure vessels. So within the epistasuction tank immersed in the tank are a series of high pressure composite bottles containing helium. The helium is used to pressurize the stage. So as the oxygen and fuel deplete, you have to replace that volume with with a gas and provide a minimum feed pressure to the to the engine and a pressure to support to structurally support the stage. it it appears that one of those struts broke free during flight. now we we do a
5:49whole lot whole bunch of tests on the ground but part of the reason why this that that test did not identify the strut is that it is that is something that's sort of quite unique to rockets is as the as as the stage is accelerating the buoyancy increases proportionate to the gloing. So, it it might seem sort of a bit counterintuitive that as the as as
6:18the rock's accelerating that the that something in most in the tank would actually want to go up more, but but that's basically what what happens is the buoyancy increases proportionate to the the G- loing. So, at approximately 3.2 gs, this this the the strut holding down the one of the helium bottles appears to have snapped.
6:42and and and as a result releasing a lot of helium into the upper stage oxygen tank and causing over pressure event quite quickly. so it should be said however that we've flown several hundred such struts on on every Falcon 9 vehicle meaning there's a cumulative flight history of several thousand of these struts.
7:13moreover the strut that we believe failed was designed and material certified to handle 10,000 lb of force but actually failed at 2,000 lb of force which is a five-fold difference. also in examining detailed close-up photos of stage construction, we don't see any visible flaws or damage of any kind. so it is part of what what caused this to be to be quite a puzzle. the the telemetry data also
7:44somewhat confusingly shows a drop in the helium pressure system which you would expect if there's a breach in the helium system and then somewhat strangely a rise in the helium system back to approximately its starting pressure. this is this was obviously quite quite confusing, but we think what may have happened is that as the the helium
8:09bottle broke free and twisted around, it may have pinched off the the the line to the helium manifold and restored pressure in the helium system but released enough helium into the liquid oxygen tank to cause the the liquid oxygen tank to fail. This is somewhat speculative, but that's that's the best explanation we can we can think of right now. So, it's a really really odd odd failure mode.
8:39so the you know so within the within the course of a second that this caused the enough helium to be released we believe to over pressure the liquid oxyuction state stage liquid oction tank of the upper stage. and you don't really need to release a lot of helium because there's only about a about 2% gaseous volume in the up in in the in the in the stage because the upper stage propellant is not being consumed. The first stage propellant is is being consumed obviously as first stage is is frosting but the upper stage is not. So, obviously what we're going to do in the future is we're we're obviously not not going to use these particular strutes. and we're going to move to
9:32individually testing each strut independent of any material certification. So, no matter what something is actually certified to handle, we're not going to believe that. We're simply going to individually unit test every single strut. and this this will result obviously in in some some cost increase to the rocket, but not we believe of a significant amount and not nothing that we would we think should affect the the actual price of the vehicle. But but nonetheless, it's it is some increased cost that we will apply to the to to the vehicle.
10:15so far the investigation is not showing any other issues. but as I said, we're looking closely for any near misses and and we intend to address anything that that we can possibly think of to improve the probability of success in the future. that this is the first time we've had a failure in in seven years. So it's I think to some degree we were a little I think the company as a whole became maybe a little bit complacent over the course of seven years and potentially
10:5420 successes in a row and but I I think this is certainly a an important lesson and something we're going to take with us into the future. so with that, I'm happy to take it turn it over to your questions. At this time, I would like to remind everyone in order to ask a question, please press star, then the number one on your telephone keypad. We'll pause for just a moment to compile the Q&A roster. Your first question comes from Eric Berger from Houston Chronicle. Your line is open.
11:29Yes. Hello. Thanks for having this this call. I guess the question I would have is do you have any estimate on when you'll be able to address the strut issue and get back to flying regular Falcon 9 launches? Yeah, I think the the addressing the strut issue is is fairly straightforward.
11:47we switching to something which has an even higher margin of of safety than than what we had which was already very high. and as I said individually testing each one. So ju just based on this on moving to like a an absurdly strong strut and and individually testing a strut that that would only push our flights out by maybe a few months. but we want to make sure that we've turned over every piece of data, that this truly is the that
12:24we diagnosed the issue correctly and that there are not some other issues that need to be need to be addressed as well. And then of course we want to make sure that all of our key customers are on board with that. so particularly NASA, Air Force, FA and our commercial customers. so we don't want to give an a precise flight date until we've until everyone is on board with with a new flight date. But just based on the strata it the strut itself does not significantly push out the next launch.
13:03Your next question comes from Alan Bole from Geekwire Cosmic Log. Your line is open. Hi, thank you for taking the question. I wanted to just get another if you could explain how the strut led to that over pressure again. did the the helium tank come loose and then somehow the increased pressure of the helium got into the got into the second stage tank. Just just if you could go over the relationship between the strut coming loose and the over pressure again, I'd appreciate it. Thank you.
13:37Sure. so the in in the liquid oxygen tank on both stages but was talking specifically about the upper stage the there are high-pressure helium bottles. These are composite helium bottles that are at about 5,500 PSI. They're stored in the liquid oxygen tank in order to chill down the helium that they contain to to cryogenic levels which improves the density of the helium considerably.
14:14and that that let me just a little bit beyond the failure. The the helium from those tanks flows to the engines or in upstate upstage case the engine where it is heated and then returned into the oxygen tank and the fuel tank to pressurize those tanks and replace the volume of oxidizer and fuel that is depleted while the engine burns.
14:42and the to provide pre pressure which structurally stabilizes the stage. so it it looks like the the strut the key strut that holds down one of the helium bottles. so se there's a whole bunch of these healing bottles. One of those struts appears to have failed. and as a result it was unable to to hold the helium bottle down. And so the human bottle would have shot to the top of the tank at high speed because of the the extreme buoyancy at the failure point of 3.2gs. now although we we we fire the
15:25stages on the ground and we do stage hold down firing at the launch pad the force the buoyancy force is only 1g of course because we're it's not accelerating. and so it would have passed through all of those tests without having an issue because it only sees high gloating in in flight.
15:50there's also it's it's very difficult obviously to test a rocket stage because the only way to fully test the stage would be have some enormous centrifuge that doesn't exist on Earth that would centrifuge the the stage up to to flight acceleration. but that that's just not it's not possible. That that would be the ultimate test. But there's no there's no such physical thing that exists on Earth that could do that.
16:17so one one is forced therefore to do these things by proxy and the proxy is sometimes inadequate as just turned out to be the case here. Your next question comes from Marcia Dunn from the Associated Press. Your line is open. What what is the size of the stress that failed? How long was it? How thick was it? I'm sorry if I missed the material.
16:44And are you considering putting parachutes on future cargo capsules? yeah, I mean I wouldn't I don't sure the exact dimensions of matter that much. I mean it's it's sort of about a about 2 ft long and about a an inch thick at its thickest point. so the the there are parachutes on dragons always. the the the reason
17:20I mean that's how it normally lands. the the issue is that in this case we didn't have the the software on Dragon is is is inert on Dragon one. the the software to initiate the parachute deployment was not in Dragon 1. it it certainly was planned for Dragon version two. and what one of the things we've decided to do is to advance the software that was meant for Dragon version 2 into Dragon version one. So that if something like this were to happen in the future, the Dragon spacecraft would save itself.
17:58Okay. Yes, that's what I meant. And will it be on the next flight? Do you think the next flight of Dragon? Yes, it will have the it will have the software to initiate parachutes in the event of a launch failure. Thank you very much. Your next question comes from James Dean from Florida today. Your line is open.
18:19Hi. thank you, Elena. Could you just please clarify? Did you say that you expect it will be at least a few months before you're ready to fly again? would which flight will that likely be? And I just wanted to also ask if this has impacted your commercial crew timeline at all. I I I mean the best I could say with respect to flying is that it's a a no sooner than September date. It could be September, but it's no sooner than September.
18:51we don't want to it's it's also not clear who that customer will be. So I can't say exactly which there may be some change of which customer it is on that flight. yeah sorry do you have another element to the question? Well I was just curious if if you know as this has delayed you know the entire manifest obviously is it also commercial timeline?
19:16I don't I'm sorry. Go ahead. Yeah, I don't think this affects our commercial crude timeline. because this doesn't affect the critical path on our commercial crew timeline which is the critical path is really the design and and validation of Dragon 2. So we're not expecting any change in that.
19:43Well, thanks. And and maybe I think Gwen spoke to this earlier, but would you affirm that you think a crew how do you think a crew would respond in a the system would respond if a crew were on the vehicle in in a situation like this? Thanks. Sure. For Dragon 2, the the system would have initiated an abort thrusters. So it would would
20:08thrust itself away from the vehicle quite rapidly and and then once it was decent way away from the vehicle would initiate parachutes to to land in the ocean. the but as I said even Dragon one which does not have the abort thrusters survived the the upper stage over pressure and we actually received telemetry all the way until just above the ocean where we we lost line of sight in the telemetry. So if even with Dragon 1 without the abort thrusters if it had initiated shoot deployment the the cargo would have been saved or a crew would have been saved conversely. so that's that's unfortunate we didn't have that software in on this mission but
20:58certainly will be there on on other Dragon 1 missions and it was always baseline to be there for a crew mission. Your next question comes from Dana Ho from Bloomberg News. Your line is open. yes. Hi. The Struts in question, were they made in-house by SpaceX or were they furnished from a supplier?
21:21the the the part that that failed was was from a supplier. but I mean, we don't we we're not going to name the supplier because I I think that just results in recriminations. But it was not it was not made at SpaceX. and we were re relying on material certification from a supplier.
21:43and so there with that you know in the future we independent of material certification we will be individually testing each threat. Okay. And it and would you ever consider just bringing that inhouse as well? I know that, you know, one of SpaceX's claims to fame is that so much of the rocketry is made in-house by SpaceX in Los Angeles. Would you consider making the struts yourselves as well?
22:13in in this case, we we don't think that the strut is something we we need to make inhouse. we are going to be long-term switching to a a different strut design. from a different supplier most likely but but I don't think this is likely to be it done in-house not not this particular component and and I should say while SpaceX does all of the the major components inhouse it there are hundreds of we have hundreds of suppliers that that supply sort of minor components and it'd be very difficult for us to insource all of those.
23:01yeah, sure. And then real quickly, in terms of debris, how much of Dragon was recovered? it sounds like it did a heroic job given given everything that went on during that event. It did. it's somewhat tragic that it unfortunately hit the water quite hard. and is sort of in at the bottom of the ocean. We were actually sending try trying to send a remote submersible down to the region where where the debris hit the water and see if we can retrieve additional elements of of debris to either confirm our initial conclusions or avoid us in a different direction.
23:45yeah, that that is probably the the saddest thing about this is that if with with just a bit of different software, Dragon would have made it. Your next question comes from the line of Irene Claus from Raiders, your line is open. have have you guys been able to replicate the failure and have you eliminated all other failure possibilities from your fault trees?
24:12Thanks. we we have been able to replicate the failure by taking a huge sample of of these essentially thousands of these struts and and pulling them and then we found a few that that pulled far below their their pificated level. and that's what led us to think that this that that essentially there was there was one that was just far below its its rated capability that that
24:48happened to be in the wrong place at the wrong time. but but as said in the beginning of the call, we we do want to characterize this as an initial assessment and and it's subject to further review of the data to eliminate any other possibilities in the in the fall tree. Right now it there doesn't seem to be any other explanation that that that makes sense. but we're continuing to investigate.
25:16Thanks. And what was the material that the strut's made out of and any rough idea of how much it cost? I don't want to get into too much detail on on on this. it like we're getting into the nitty-gritty. and and it sort of invites like a lot of armchair analysis that that just isn't actually correct. So I mean just broadly speaking it's made of steel but and like like I said is not something that should ever have failed at this force level.
25:51but it would appear to be incorrectly made but with no no visible way of determining that from the outside. So thanks Elon. [Music] your next question comes from Frank Morig from Aviation Week. Your line is open. Thank you. could you tell how was there just the one strut holding the the the tank or how how it fit into the into the structure of the the locks tank? Were there more than one? And where on the strut if you could say where where did it fail or do you know?
26:32Yeah, we we have a pretty good sense for that that that I mean this there's one strut sort of holding it in the vertical direction. I said that the strut is is is oversized to be able to handle far in excess of the projected load. and you know in this case it it failed far far below it you know it failed five times below its nominal strength which is which is pretty crazy.
27:05and you know yeah but appeared to fail fail at the bolt head most likely. there's a steel bolt head appears to have failed there. we think u but but it's sort of this it's just it's hard to get into the nitty-gritty because then it makes it sound like far more definitive than than it is.
27:34in your with further investigation turned out to be something else. Yeah. In in your redesign, would you would you are you adding another strut or more than one somehow to strengthen it beyond what you had this time? I I we don't think that more than one strut is makes sense. It just needs to be a strut that it just needs to be that each strut is is going to be individually tested independent of its material certification.
28:08I mean the the the strut is meant to handle more than twice the actual load that it experiences. So you could in principle divide that into into two struts, but you're not really gaining anything if you do that. Thanks. Your next question comes from Bill Hward from CBS News. Your line is open. You is it just the one helium tank Elon or is there more than one? And and I'm I'm curious based on what you just said, is there is there actual data that indicates a strut broke versus one of the helium overwrap tanks, you know, bursting on its own? How how can you differentiate that sequence if if you would please?
28:56Sure. and you know, again, at the risk of of of being a bit repetitive, I do want to emphasize we are not characterizing this as a definitive result. we're characterizing this as a preliminary result. as the as the most probable outcome, not as a definitive outcome. the yeah so there there are
29:24several healing bottles in in the upstage oxygen tank. It actually depends on on some missions have more healing bottles than others. but there are a whole bunch of them. the the the the reason we don't think that this was I mean there are there are a few reasons and again these are just things that make it more probable rather than less probable is that
29:55if we look at acoustic triangulation so if we have we've got microphones or technically accelerometers at various points on the upper stage. And by looking at the exact timing of high frequency events in the on the stage, we we can by acoustic triangulation identify the location where we where the the snap occurred or or or where the breakage occurred because it's sort of sort of a obviously would be a sound.
30:34and the acoustic triangulation points to the support strat as the most probable location. also if let's say the the helium bottle liner broke the the the composite overlap is is not going to break. But the if there was was say a crack in the metal liner of the helium bottle it would have resulted in a continuous release of helium and you would have seen much more helium release than if if if the bottle had simply broken free and and could thus twist sort of twist tie closed the line that connects it to the rest of the stage which would which would explain why the helium pressure initially drops and then and then and then restores.
31:29Thank you. Your next question comes from Jeff South from Space News. Your line is open. Yeah. have you briefed various customers insurers about your preliminary results so far and what sort of feedback or reaction have you gotten from them so far? Thanks. yeah, absolutely. you know in in deference to our customers and and
31:55and the government, we we made sure to brief them last week and before briefing the public today. and so they they've looked at all of the same data that we have. they have complete access to the data center that we have limited only in the case of international customers by ITAR and and they they've they agree with our conclusions thus far again emphasizing that they are preliminary and we've we need to continue working through the full tree and but but I I mean I'm happy to say that every one of our customers has and supportive and none of them have indicated
32:43a diminished faith in SpaceX. so I would just like to you know say give a word of of appreciation for the support of our customers. Your next question comes from Scott Powers from Orlando Sentinel. Your line is open. Thank you. I'm wondering if you have the same strut and helium tank configuration on the lower stage or something similar to it.
33:11There there is it's it's the same strut on the upper and lower stage. So you'll be replacing it in both stages. That's correct. Yeah. Was there any is there any risk this the strut the strut is slow on many many prior flights that it's Yeah. Yeah. Was is there any risk that the lower stage struts could have reached the same G levels? And I guess they would have already done that at that point. Yeah, they were all at the same same G levels. now the the buoyancy force once the the liquid level drops below the the position of the helium bottle, the buoyancy force drops dramatically, of course. but but but that was not the case at that point.
33:54the that all the helium bottles were experiencing approximately the same buoyancy before we both upper and lower stage. [Music] Your next question comes from Dan Verano from Buzzfeed. Your line is open. Thanks very much. you'd said the shame of this was that the Dragon parachutes weren't enable the software to fire.
34:19I just wondering could you why weren't they? Was that just too much trouble or was there a decision that went into that? well, with with the with the version one, the cargo dragon, it was not it was just not something that had been done yet. So, it was it was on the agenda, but it hadn't been done yet.
34:43and it was certainly planned for Dragon version two for both cargo or crew. and so it was planned to be done. It just had not been done as of that flight. It will be done as of the next flight. Thanks very much. Your next question comes from Mike Wall from space.com. Your line is open.
35:07A lot for for actually doing this. yeah, I mean like you mentioned that that like you guys may have gotten a little complacent like over that long run of success. Could you just go into a little detail about I mean how do you think that sort of manifested itself and so how is this incident going going to change your guys' outlook going forward? sure. Well, you know obviously we had you know several years ago in the early days of Falcon 1, we we when we were just sort of learning how to make rockets, we we had some challenges there with the first Falcon 1 flights. and I think that that instilled an extreme level of paranoia in the whole team. But you
35:47know 7 years ago, which is when we had our last failure, the company was only about 500 people. and now we're 4,000 people. So the vast majority of the people at the company today had only ever seen success. and when you've only ever seen success, obviously, you don't feel you you don't fear failure quite as much. and you
36:14know now now before every flight I I always send out an email companywide saying if anyone can think of anything any possible reason to to hold off on on launching that they should they should call me immediately on my cell phone or send me an email and whether their manager agrees with it or not. but that's sort of something I've always sent before every flight. But, you know, the 20th time I send that email, it it just seems like, you know, there's Elon being paranoid again. So, maybe it doesn't
36:53it doesn't resonate with with the same force, but I think now everyone at the company appreciates just how difficult it is to get rockets to orbit successfully. And I think you know we'll be the stronger for it. Thank you. Your next question comes from Ken Kramer from Universe Today. Your line is open. Hi. Thank you. Thank you for doing this. I was wondering about the struts. I understand you don't want to talk about the material of construction, but you said it was stainless steel. I wonder can you say would would you at least change that material maybe maybe maybe to a different alloy? I'm also wondering about the the the the abort thrusters. Might you consider installing
37:40abort thrusters on this cargo or using even an early version of Dragon V2 at some point to do the cargo deliveries? Thank you. Yeah, we are we are likely, you know, this is still under review, but we are likely to change the material in the in the bolt of a sport strut. most likely to incanel but that's again still preliminary. and and then with respect to the the V thrusters, the V thrusters are really quite a radical change in design. and it it wouldn't be easy to retrofit those Dragon one. it it's really it
38:31Dragon 2 is very different from Dragon 1. It sort of looks maybe sort of similar from an external standpoint, but but there's no reasonable way to to just sort of tack on the aboard thrusters. the best thing to do is is to try to get to get Dragon 2 to flight at the earliest opportunity because it is it is a step function improvement in capability and and flight resilience. although as as I said, we are going to make the software change to to Dragon. Dragon is quite a a tough tough creature. even Dragon 1 and as I said it it it actually was okay even after the upper stage explosion.
39:16it was broadcasting it telemetry all the way to when we lost line of sight just above the ocean. so it's quite tough and even without the aboard thrusters like would have made it. So really it's just a question of you firing the the shoots and and so we're going to we're we're going to make sure that before we fly Dragon one again that it has that ability.
39:44Thanks. But is is there any chance that you would fly an early version of Dragon 2 as a cargo Dragon? The the the first Dragon 2 will be without astronauts. So the at least one flight maybe two flights will be without astronauts. so it would the first one would in fact carry K cargo.
40:05Okay. Thank you. Your next question comes from Steven Clark from Space Flight Now. Your line is open. Hi Elon. Thanks for taking our questions. a couple of questions. you mentioned you've tested these struts on the ground and some of them failed. Did they fail at sort of the 2,000lb force mark or or where do they fail?
40:26What sort of loads do they fail under in these ground tests? And also with a few months delay in your manifest, do you expect the first flight of Falcon Heavy will probably slip into next year now? Thanks a lot. yeah. at first we didn't think it was the strut because we tested a whole bunch of struts. None of them failed at at
40:55at a level like the lowest we saw anyone fail was at 6,000 lb of force. So first in the initial part of investigation where we tested a whole bunch of these struts, none of them failed below 6,000 lb. So we thought well that that that couldn't be the issue. So we started looking elsewhere couldn't find any issues you know couldn't find an explanation for the data any anywhere else. So then then we got a just a huge number of of of these struts with this particular bolt.
41:28and after testing some enormous number I mean we were able to find one that that that failed below the 2,000 lb level. or yeah so it it was sort of a statistical thing. and then we we did some material analysis on that and found that the the there was there was there were problems with the the grain structure of the steel. It it hadn't been formed correctly. So so we think that that probably was was a bad vault that slapped through that that that looked looked good looked good but wasn't actually good on the inside.
42:11Your next question comes from the Falcon Heavy. Oh yeah, right. The Falcon Heavy. yeah, I think G given our focus on on Falcon 9, we've dep prioritized Falcon Heavy to probably launch in the spring next year. So maybe April or so. Okay, your next question comes from Tim Fernho from Courtz. Your line is open. hi guys. Thank you for taking time to do this. just two questions.
42:49can you describe the insurance coverage that you had for that rocket and for Dragon and its cargo? And has there been any cost SpaceX has had to pay because of delays to its commercial launch contracts? no. The the the launches we don't insure our launches. it's it's actually quite I think nobody in the rocket business does except for potentially third party damage on the ground which is a super low probability thing. so it was no no launches
43:18there. the the biggest penalty is SpaceX will be the delay of launch rate. So you know if sort of like an airline if if the flights don't take off then you lose the revenue associated with that period. so it the the lost lo last revenue will be you know meaningful I mean in the hundreds of millions probably because of the implied delays.
43:46Thanks. Your next question comes from Joe Papalardo from Popular Mechanics. Your line is open. Yeah. Hi. Thanks for for taking the call. most of my questions are have already actually been been addressed, but I wanted to know in particular if you had recovered anything from from the from from the debris that that helped at all or and you said you had put in a an RV, a submersible. Was that what do you have any details on that and anything recovered from those efforts? Thanks.
44:22We did find a small amount of debris that was floating. most of it's at the part of the ocean. The remote submersible. we haven't gotten the feedback from that yet. I think it's it's going out soon, but but it's it's a lot of area to search. it's pretty dark down there. So, I wouldn't give we're just doing that as a as sort of because we want to do everything we can think of.
44:51and and if if we do find something obviously we'll feed that into our failure investigation but but but so far we've not I don't think it's actually reached the area in question. so we would yeah so we just have a small amount of floating debris. That's better. Your next question comes from Sophia Chen from Wired. Your line is open. Hi thanks for doing this. I just wanted to ask, I had two questions. one was, so you mentioned the strut, is supposed to be able to take 10,000 lbs.
45:27like what what is, like how much force do you actually need? And then my second question is, between these, successful launches and this particular failure, like what how how has the design of Falcon 9 changed or if at all? Yeah. So the the the actual force experienced in flight is proportionate to the g- loing. So since it failed at 3.2 gs the and you know really the max gs is that it experienced depending on mission is maybe 5 to 5 1/2. the the most amount of force would really experience would be maybe 3 to 3,000 lb of force. So it sort of you
46:16know theory was failing at three times below if it if it had reached the the the worst case at the end of flight it would have been 35,000 lb of force but still yeah like a three times below its nominal rating and yeah I think the way that this what what we'll be doing going forward is doing a lot or individual unit
46:44tests. So essentially pull tests on on on on all all the supporting hardware. so that even if something is material certificated for for very high loads, we that's nice, but but we're not going to believe the certificate. We're just going to we're going to test it as well. Your next question.
47:15Yeah. My my next question was how has the how has the design of Falcon 9 how is it different at all this time compared to the successful launches of the past? I'm sorry I'm having trouble hearing you. Is this design different from Is this design different from prior successful missions? Yes. Is that Is that what you're asking? no. It's the design is identical to to prior missions. So, we've already we've flown hundreds of these of of stress of this exact design.
47:49It's actually thousands. your next question comes from Graham Star from the Christian Science Monitor. Your line is open. I just want to ask if you found that any of this was due to the way it was implemented by the SpaceX team or if you found it was purely material and regardless of what you found, you've talked about focusing on individual tests of the struts. U, but does that change anything regarding future quality assurance or operations policies?
48:20yeah, I mean this this was a purchased part. so it's not it's not really related to anything. I mean we just install it at SpaceX. so and we have the closeout photos from this particular mission. So even with looking at detailed high resolution closeout photos the it it appears perfect on the closeout photos. so there there does not appear to be any assembly error. we should should be able to see from the the closeout photos.
48:55and yeah the the the big thing is the big sort of policy change would be or always that pull testing or or compression testing even if it's a supporting even when it's a supporting component that that in theory has extremely high margins and has certificates to to sort of allegedly prove that the The policy change would be ignoring the fact that it has a certificate and and and and in theory very high design margins and applying a physical test to even the the small support items of support support hardware. We we already do that for the macro items for the stage the the primary airframe is is always tested.
49:51but but we we had not applied that to most the supporting hardware particularly where the the expected margin of safety was extremely high. Your last question comes from Sean Okaine from the Verge. Your line is open. Hi. Thanks for taking our questions. I know you talked about potential Falcon Falcon 9 delays and Falcon Heavy delays.
50:18I'm curious if you think this will affect the second commercial crew or cargo contract being awarded in the fall by NASA. I don't know. I I don't think so. But that would be a question for NASA. I mean your submission for that. Oh, our submission. it wouldn't affect our submission now.
50:45I mean it has been sort of I mean just emphasizing the difficulty of rockets in general obviously I mean just just for space station resupply as I think most people on the phone know I mean the last year has seen three failures. So there was ours orbital sciences and the Russian progress. So
51:04it's it's been a been a tough year and each one of those failure modes has been quite different. So I mean it just goes to show like rockets are a fundamentally difficult thing. I mean generally when there's an idiomatic expression about something there's a reason for it. We have no further questions at this time. I turn the call over to the presenters. All right. Thank you.
资源与出处
- 视频在 YouTube 收看
- 档案silicon-valley.tech 档案页本页视频与文稿数据来自 Elon Musk Archive
- 精读本篇精读 Markdown 版