Launches fail when one of thousands of parts or decisions goes wrong at high speed — a motor losing pressure, a leaking valve, a misread sensor or a shield that stays shut.
A rocket is a long chain of events that must all work, in order, within minutes. Thousands of parts are pushed close to their limits. One weak link is enough to lose the mission. India's record shows the main ways it happens. In each case below, a single part or signal — a separation mechanism, a valve, a sensor reading, a motor — was enough to end the mission.
Sometimes the rocket works and the satellite fails. In January 2025 GSLV-F15 placed NVS-02 correctly in its transfer orbit. Then a command signal did not reach a valve on the satellite's engine, and it could not climb to its working orbit.
ISRO sets up a Failure Analysis Committee. It studies flight data, runs computer simulations and repeats the suspect conditions in ground tests. After GSLV-F10, the committee found the most likely cause — a leaking vent and relief valve — and recommended a new tank-pressurisation system, a stronger valve and extra automatic checks before lift-off. After SSLV-D1, the separation system, sensor logic and deck structures were all changed. SSLV-D3 placed its satellite precisely in orbit in August 2024.
Read the 2025–2026 entries for PSLV-C61 and PSLV-C62, two failures in the same third stage. Explore PSLV's history
Why reliability is hard. If a launch has N critical events and each succeeds with probability r independently, the mission succeeds with probability r^N. With r = 0.999 and N = 100 events, the chance of success is 0.999^100 ≈ 0.905. Even 99.9% reliability per event leaves about a 1-in-10 chance of losing the mission. Engineers push r up with testing and margins, and cut the effective N with redundancy.
Redundancy and its limits. If two independent backup paths each fail with probability p, both fail with probability p². For p = 0.01 that is 0.0001. But redundancy only helps if failures are independent. On NVS-02, ISRO's committee found that the drive signal failed to reach the pyro valve, with the most likely cause a lost contact in a connector — and the problem affected both paths. A shared weakness defeats the backup.
Reading a fault chain: GSLV-F10. ISRO's committee laid out the sequence:
The pattern — symptom, mechanism, root cause, confirmed by simulation and ground test, then fixes — is how every failure review works.
Same stage, twice. PSLV-C61 and PSLV-C62 both had trouble in the solid third stage. A repeated failure in one subsystem is a strong signal that the cause is in its design, production or acceptance testing rather than chance.
0.999^100 ≈ 0.905, or about 90.5%.
A leak in the vent and relief valve of the liquid hydrogen tank, which left the tank under-pressurised so the fuel booster turbopump fed too little hydrogen to the engine.
No. GSLV-F15 placed NVS-02 correctly in its transfer orbit. The satellite's own engine could not fire because the drive signal did not reach an oxidiser-line valve.