ERA21LA173 · Piper PA-23-250

The instructor and pilot receiving instruction were conducting multiengine flight training. During takeoff following a simulated engine failure on the runway, the left engine lost power, and the instructor stated that the left propeller would not feather. The airplane descended into terrain past the end of the runway. Witnesses stated that their attention was drawn to the airplane due to its “unusual” sound that was inconsistent with takeoff power. One witness said he could not discern if one engine or both engines were making “continuous sputtering/backfiring” sounds. The airplane climbed to about 100 ft above ground level and the landing gear remained extended until the departure end of the runway. Shortly thereafter, the airplane entered a shallow turn to the left until it disappeared behind a tree line.

Automatic dependent surveillance-broadcast (ADS-B) data revealed that the airplane achieved a groundspeed of 86 knots about midfield and slowed once off the ground. About 200 ft agl, the track depicted a descending, decelerating turn to the left. The radius of the turn tightened until the last target was recorded in the vicinity of the accident site, about ground level, at 59 knots groundspeed. Based on the estimated point at which the takeoff started, the airplane was over 1,400 ft into the takeoff roll when it became airborne. Performance information in the Owner’s Handbook for the airplane indicated a 750-ft takeoff distance. Although ample runway remained on which to safely reject the takeoff, the instructor allowed the pilot to continue the takeoff despite the excessive distance required to become airborne and the loss of left engine power. Pilots who had flown the accident airplane during the week before the accident described the left engine either stopping or running roughly with the fuel selector in the left inboard tank position. When the fuel selector was moved to the left outboard tank position, the engine could be restarted, or smooth, continuous operation would be restored. Each said that these power-loss events were reported to maintenance for correction. Three days before the accident, a flight instructor could not start or sustain power on the left engine with the inboard tank selected but started and ran the engine continuously on the outboard tank. He then demonstrated the discrepancy to company maintenance personnel before he rejected the airplane for his scheduled flight. Examination of the wreckage revealed a 12-inch length of duct tape, employed as a “gasket” to seal the loosely fitted left inboard fuel cap, unsecured inside the fuel tank, where it likely blocked the fuel supply port on the accident flight, as it had intermittently during the days before the accident. Examination and testing of the airframe, engines, and components revealed no evidence of any other preimpact anomaly that would have prevented continuous engine power; however, these examinations and a records review revealed numerous examples of maintenance work that was incomplete, inadequate (including the use of duct tape on the left inboard fuel cap), or not performed; the recommended engine and propeller overhauls were more than a decade overdue.

Verbatim NTSB analysis and probable cause from the NTSB dataset