On November 3, 1995, about 1102 eastern standard time, a Boeing B-17G, N3703G, registered to and operated by Military Aircraft Restoration Corporation, collided with the localizer building at the Fayetteville Regional/Grannis Field Airport, Fayetteville, North Carolina. Visual meteorological conditions prevailed at the time and no flight plan was filed for the 14 CFR Part 91 business flight. The airplane was substantially damaged and the commercial-rated pilot, airline transport-rated copilot, and commercial-rated other crewmember were not injured. The localizer building was damaged by the left wing of the airplane. The flight was originating at the time of the accident. The copilot who was taxiing the airplane to takeoff stated that pressure gauges in the cockpit showed adequate pressure and he operated the brakes without incident. The airplane was about 100 feet from a 90-degree turn to the threshold of the runway when he observed that the brakes had failed. The pilot then elected to secure the engines and he noted that the hydraulic pressure gauges showed no pressure. The airplane continued on the taxiway down a slight grade and the pilot then applied full throttle to the No. 1 and 2 engines. The engines responded momentarily and the airplane rolled off the right side of the taxiway. The pilot then elected to restart the No. 1 engine and the airplane rolled across the threshold to the runway down an embankment. The pilot applied full power to the No. 1 engine which began turning the airplane to the right but the left wing of the airplane then collided with the localizer building. Postaccident inspection of the hydraulic system revealed no evidence of leakage under static conditions or when the brakes were applied using the parking brake. A hand pump located near the copilot's seat which is used to supply pressure for ground servicing operations or to recharge the accumulators due to failure of the electric pump was operationally checked. The pressure check valve was determined to be intermittent but after 25 strokes, the pressure gauge indicated over 900 psi which held for 15 minutes. The hydraulic pressure was then reduced and the hand pump operationally checked three more times with no further discrepancies noted. Operational check of the electrically operated hydraulic pump revealed that it built the system pressure to 1,000 psi then shut off automatically. The parking brake was then applied with no drop in pressure noted. A switch in the cockpit which when placed from the auto position to the manual position maintains hydraulic pressure without an automatic shutoff was operationally checked. No discrepancies were noted. The hydraulic pressure was then released and the electric pump was activated several times and after the second attempt, the system pressure would not increase. Attempts to increase hydraulic pressure using the hand pump were also unsuccessful. The pressure relief valve was then tapped and an audible noise was heard. The electric pump then increased pressure and tested satisfactory; however, the hand pump failed to increase pressure again.