When a Slender Giant Stumbles: Long March 7A, the 85-Second Anomaly, and the Physics of Failure

当修长巨人踉跄：长征七改、85 秒异常与失败的物理法则

2026-08-10, 20:02 CST, Wenchang LC-2.​ A Long March 7A (CZ-7A / 长七改) carrying the ChinaSat-4B communications satellite lifted off cleanly, rode through the first minute of flight on a nominal thrust curve—and then, at T+85 seconds, in the first-stage burn just before booster/core jettison, it came apart in a fireball over the South China Sea. Xinhua’s dispatch was three lines long: rocket flight abnormal, mission failed, cause under investigation.

2026 年 8 月 10 日 20:02（北京时），文昌二号工位。长征七号改（CZ-7A）托举中星 4B 通信卫星起飞，前 80 秒推力曲线正常；随后在 T+85 秒、一级飞行段、助推器/芯一级集束分离前突然解体，火球坠入南海。新华社通稿只有三行：“火箭飞行异常，发射任务失利，具体原因正在进一步分析排查。”

It was not a “year of failures” for the Long March 7 family. It was one failure inside a string of successes: CZ-7 (baseline) flew Tianzhou-10 in May; CZ-7A flew three clean GTO missions in May, June, and July. The August 10 event was the 2nd failure in 18 flights​ of the CZ-7A variant, and the first since its maiden flight blew up in March 2020.

这不是“长七今年总是失利”。而是成功序列里的一次断裂：基线型长七 5 月打天舟十号成功；长七改 5、6、7 月连续三次高轨任务成功。8·10 是长七改 18 飞中的第 2 次失败，也是 2020 年 3 月首飞失利后第一次再摔。

1. Two rockets behind one name | 一字之差，两型火箭

The “Long March 7” label hides two different vehicles.

“长征七号”这一个名字底下，其实是两型火箭。

CZ-7 is the ISS-cargo truck. CZ-7A is the high-orbit specialist: it bolts a Long March 3A-series hydrogen third stage onto the CZ-7 lower stack, trades LEO muscle for GTO reach, and pays for its elegance with a slender, flexible airframe​ that behaves less like a rigid spear and more like a tuned beam under load.

长七是空间站“货运专列”（LEO 200×400 km）。长七改是把长三甲系列氢氧三子级嫁到长七下半身上，换 LEO 力气换 GTO 运力（>7 t），代价是长细比 >18、国内第一修长——飞行中不像刚性长矛，更像一根受载的梁。

That slenderness is not cosmetic. At Max-Q​ (roughly T+60–90 s, dynamic pressure peak), the rocket is simultaneously fighting aerodynamic bending, engine gimbaling authority, and a control system trying to keep a 60-meter whip straight in high-altitude wind shear. The 85-second breakup sits exactly on that cliff edge.

修长不是好看，是物理。到了 最大动压点 Max-Q（约 T+60–90 s），箭体同时扛气动弯矩、发动机摆动纠偏、高空风切变——控制系统要把 60 米长鞭捋直。85 秒解体，正好踩在这道悬崖边上。

2. What we know about the 85-second event | 85 秒那一下，已知什么

From open sources, not speculation:

公开信源（非臆测）能确认这几条：

• Timeline​ — Nominal lift-off 20:02:xx; anomaly at T+85 s; first-stage burn, pre-booster-jettison; no orbit.

• Debris​ — Low-altitude disintegration, fragments into sea; Jonathan McDowell’s read of public video agrees it never reached staging velocity.

• Official line​ — “Flight abnormal, cause under investigation.” No engine named, no structure named, no FTS (flight termination system) confirmed.

• Pre-text context​ — CZ-7A had 16 consecutive successes since the 2021 return-to-flight; the baseline CZ-7 has never failed in 2026.

时间线：20:02 起飞，T+85 秒一级段、集束分离前异常，未入轨。残骸低空解体落海。官方只说“异常”，没点发动机、没点结构、没确认是否触发自毁（FTS）。背景是长七改自 2021 复飞后 16 连胜，基线长七 2026 零失败。

Everything beyond that—pump cavitation, structural buckling, control divergence, FTS trigger—is a hypothesis bracketed by the video and the flight phase. The CZ-7A telemetry bandwidth (6 Mbit/s Ka-band via TDRS) means the data to close the case already exists on the ground; we just don’t have it.

再往外走——泵空化、结构屈曲、控制发散、自毁触发——都只是按视频和飞行阶段反推的假设。长七改有 6 Mbit/s Ka 天基遥测，判案数据其实已经在地面，只是没公开。

3. Ghost of 2020: cavitation, not recurrence | 2020 的幽灵：空化，但不是复发

The last CZ-7A failure (2020-03-16, maiden flight) was not​ a Max-Q breakup. Telemetry showed a clean curve to T+162 s, then collapse. Root cause, closed in ~20 days by CALT’s fluid-simulation team: cavitation at the booster LOX tank outlet, which depressed pump inlet pressure below the YF-100 engine’s floor and starved combustion.

上一次长七改失败（2020-03-16 首飞）不是​ Max-Q 解体。遥测飞到 T+162 秒曲线都漂亮，然后塌了。一院用 20 天闭环定位：助推器液氧箱出流口空化，泵入口压力掉到 YF-100 工作下限以下，烧不起来。

That fix—reshaped outlet, cavitation-suppressed feedline—flew successfully on every CZ-7A mission from 2021 to 2026-07-29. So the 2026 event is unlikely to be the 2020 bug reborn; the phase (T+85 s, Max-Q, pre-separation) doesn’t match the 2020 signature (T+162 s, upper-first-stage transition). Treating them as “the same failure” is the lazy reading.

那个改法（重设计出流口、抑空化输送路）从 2021 用到 2026-07-29 全部成功。所以 2026 这次不太可能是 2020 老毛病复活——阶段不对（T+85 秒 Max-Q 分离前 vs T+162 秒一级末），硬说“又栽在同一条沟里”是偷懒。

The honest shortlist for 2026-08-10:

2026-08-10 的诚实清单：

1. First-stage propulsion anomaly​ — YF-100 family (booster or core) turbine/pump/plumbing failure causing over-pressure or uncommanded shutdown. Plausible, but the video shows upper-stack rupture before a pure engine fireball, which doesn’t fit cleanly.

2. Structural / aeroelastic failure​ — L/D>18 beam under Max-Q bending + wind shear; interstage, booster-core bundle joint, or forward fuselage buckling. The “whip” problem CALT designers explicitly called out in 2021.

3. Guidance / control divergence​ — Gimbal rate saturation, IMU fault, or software edge-case in the high-L/D attitude law, letting aerodynamic load tear the stack.

4. FTS activation​ — Range safety sending the destruct command because the vehicle was already outside corridor; explosion is then consequence, not cause.

5. 一级动力异常（YF-100 系涡轮/泵/管路超压或熄火）——但视频是先上部断裂再爆，不像纯发动机火球。

6. 结构/气动弹性失效——L/D>18 梁在 Max-Q 弯矩+风切变下，级间段、助推-芯级集束接头或前体屈曲。这正是 2021 一院设计师点名的“长鞭”难题。

7. 制导控制发散——摆动饱和、IMU 故障、高长细比姿态律软件边界，气动载荷把箭撕了。

8. FTS 自毁——安控判定出走廊主动引爆，爆炸是结果不是原因。

Until the accident board speaks, all four sit at equal weight.

在调查组开口前，四条权重相等。

4. Why “always failing” is the wrong frame | 为什么“总是失利”是错帧

Launch cadence in 2026 (CZ-7 family):

2026 长七系列发射节奏：

• May 11 — CZ-7 Y11 → Tianzhou-10 ✅

• May 27 — CZ-7A → Comsat Tech Test 24 ✅

• Jun 23 — CZ-7A → Comsat Tech Test 26A ✅

• Jul 29 — CZ-7A → Tianlian-3 01 ✅

• Aug 10 — CZ-7A → ChinaSat-4B ❌ (T+85 s)

Five attempts, four successes, one failure. The failure is visually spectacular (night fireball, 85 seconds, GTO payload), so it crowds out the four greens in public memory. But cadence-wise, CZ-7A’s 17/18 lifetime success rate​ is still healthy for a high-L/D cryogenic variant in its sixth operational year.

五发四成。失败之所以像“总是败”，是因为夜空火球、85 秒、高轨通信星，视觉权重碾压那四次绿勾。但按频次算，长七改 lifetime 17/18​ 在修长低温改型、服役第六年这个语境里，依然健康。

The real story is not “Chinese rocket keeps failing.” It is: a high-energy-density, maximum-slenderness liquid stack flying in a dense 2026 cadence hit a Max-Q anomaly the telemetry will explain in weeks.​ That is how rocket science actually iterates—not by myth, by regression.

真正的新闻不是“中国火箭又炸了”，而是：高密度发射下的高长细比低温捆绑箭，在 Max-Q 撞上一次异常，几周后遥测会给出回归结论。火箭科学就是这样迭代的——不靠神话，靠归零。

5. The closure loop: Chinese-style “归零” | 收口方式：归零

One thing the 2020 case taught and 2026 will repeat: the CALT closure path is back-to-back simulation + scaled rig test + full telemetry reconstruction, on a timeline of weeks to a few months. The 2020 cavitation fix took 361 days to return to flight; the report itself landed in ~20 days. Expect the same shape now: silent investigation, then a narrow technical bulletin, not a press spectacle.

2020 给 2026 的模板是：一院闭环 = 背靠背仿真 + 缩比试验 + 全遥测重构，周期几周到几个月。2020 年空化定位 20 天出、361 天复飞。这次形状一样：先沉默调查，再出一份窄技术通报，不是发布会秀。

If the root cause is structural/aeroelastic, the fix will be local—stiffener, interstage joint, gust-response gain, maybe a small GNC gain schedule change for Max-Q. If it is propulsion, a feedline or turbopump batch issue gets isolated. Either way, CZ-7A doesn’t disappear; it absorbs the 85-second anomaly into its next 60.1-meter flight.

若根因是结构/气动弹性，改法是局部的——加筋、级间接头、阵风响应增益、Max-Q 段 GNC 增益表微调。若是动力，就隔离某批输送路或涡轮泵。无论哪种，长七改不会退场，它会把 85 秒异常吸收进下一发 60.1 米的飞行里。

Bottom line | 一句话

The Long March 7 family is not stumbling through 2026—it stumbled once, at T+85 seconds, in the most physically hostile second of a slender rocket’s life. The 2020 ghost was cavitation and is closed; the 2026 ghost is unnamed and sitting in the telemetry. Wait for the regression. Don’t narrate the fireball as the whole story.

长七系列不是 2026 年一路摔——它只在 T+85 秒、在修长火箭一生中最敌意的哪一秒摔了一次。2020 的幽灵叫空化、已闭环；2026 的幽灵没名字、躺在遥测里。等归零报告，别把火球当成全部剧情。

— Written from open-source telemetry, Xinhua dispatch 2026-08-10, CALT post-2020 review, and Max-Q physics. No classified input.

继续阅读：长七   火箭   长七改

此文由 怡心湖 编辑，若您觉得有益，欢迎分享转发！：首页 > 赞·中华 > 经验谈 » 当修长巨人踉跄：长征七改、85 秒异常与失败的物理法则