By Joe Stephens and Lena H. Sun
Washington Post Staff Writers
Sunday, September 6, 2009; A01
In a tunnel below the Potomac River four years ago, Larry Mitchell was at the controls of a crowded rush-hour Metro train headed to Rosslyn when he saw a glimmer of red reflecting off the walls. The train's crash avoidance system indicated that the track ahead was clear, but Mitchell sensed danger in the distance. He decided to override the system and brake manually -- then watched helplessly as his train rolled to a stop just 35 feet short of a train ahead.
As a shaken Mitchell radioed Metro supervisors, he was interrupted by the operator of the train behind him, who announced that he had just caught sight of Mitchell's train and hit his emergency brake. "You could hear the panic in his voice," Mitchell said. That train ground to a halt 20 feet short of Mitchell's.
The outlines of the 2005 near-miss -- the first of three known breakdowns of a crash avoidance system designed to be fail-safe -- were made public shortly after it occurred. But newly obtained records and interviews detail just how close the trains came to what documents said would have been "disastrous collisions."
They also illuminate similarities to the June 22 Red Line crash that killed nine people near Fort Totten as well as to a March 2 incident in which two trains came "dangerously close" on Capitol Hill. The Washington Post first reported the March incident last month.
The National Transportation Safety Board, which is investigating this summer's deadly crash, confirmed that the 2005 incident has become a focus of its probe and that its investigators recently examined records from both near-collisions. They also tested hardware taken from the 2005 incident site to compare with similar equipment recovered from the crash.
Records and interviews indicate that Metro engineers did not perform exhaustive on-site tests of all components related to the incident in 2005 because they thought they had found the problem and did not want to further inconvenience passengers. Records also show regional safety officials were not formally notified that Metro had put into effect its own recommendations on how to make the subway safer.
Metro officials said they responded appropriately to the 2005 incident, identified the problem with the crash avoidance system and fixed it. A Metro spokeswoman said the cause was an electrical short circuit in cables under the tracks. She described it as "very different" from the fluctuating track circuit suspected of causing the recent accident. Cables are one component of a track circuit, which is a key part of the safety system.
Metro officials say the system is safe. Until June, no passengers had died in a crash since 1982.
After the 2005 incident, Metro's safety office made six recommendations aimed at avoiding a recurrence. By the time of this summer's crash, records show, none had been formally implemented and approved by the Tri-State Oversight Committee, which monitors Metro safety.
"We asked Metro for information on how those six recommendations would be implemented," the committee said in a statement to The Post. "We have not received any evidence that Metro has yet put them into practice."
Metro officials initially told The Post that they had put the recommendations into effect and notified the committee of the corrections years ago. In response to additional questions, a spokeswoman issued a second statement saying "we were mistaken" in asserting that the committee had closed out the issue. Nonetheless, she stressed, the corrections had been made.
The Post obtained documents related to the incidents through open-records requests to the Tri-State Oversight Committee. Metro officials did not respond to similar requests.
"I was assured that it would be looked into and taken care of," said Mitchell, who can recall coming so close to the train's red taillights that he could see the expressions of a man and a young girl in the last car.
"Four years later, I've never received any information. I have no idea to this very day what happened."'They Didn't Do It Thoroughly'
Rail safety specialists who reviewed documents obtained by The Post said there are striking similarities between the 2005 incident and this summer's crash, and they questioned whether Metro had done enough to determine all possible causes of the earlier near-collision.
After the 2005 incident, documents show, engineers concluded that a cable in a track circuit at the site failed, causing a train to "disappear when it occupied a certain position within the track." A malfunctioning track circuit is also suspected of being at the heart of this summer's crash, in which federal investigators say Metro's automated system failed to detect a stopped train and did not send a stop command to an approaching one.
Investigators have not pinpointed the cause of the circuit failure at Fort Totten but are testing a type of equipment known as a module. For comparison, they also have decided to test the same type of equipment removed from the site of the 2005 incident, said Bob Chipkevich, head of the NTSB rail investigation office.
After the 2005 incident, Metro engineers spent days looking for a cause before concluding that the problem was a short in a cable. To eliminate train delays during the investigation, they replaced the cables and other components that make up the track circuit. A Metro engineer noted in a report at the time that the investigation had been "inconveniencing patrons for a week already." He added: "The commitment to return the track circuit to normal as soon as possible precluded additional testing."
Engineers from the company that made the equipment agreed with Metro in a letter that the failure probably resulted from a short circuit in the cables. But they wrote that they "could not confirm that this was the actual root cause" because Metro separated the cables "before testing could take place."
Metro engineers did not examine the modules more extensively at the time because they were determined not to be the cause, said Metro spokeswoman Lisa Farbstein. They planned to inspect them later in a laboratory, the report said.
A couple of months later, engineers examined the modules for anything that could cause an electrical short and found no problems, Farbstein said. She said Metro has no record of that inspection, but an engineer told her that he did it.
Russell Quimby, who retired in 2007 after 22 years at the NTSB as a rail safety engineer, criticized Metro for not conducting more testing.
"They didn't do it thoroughly," Quimby said. "Anything that could have caused that system to fail, they should have tested at the time. They said, 'This is it, shut your lunchboxes, let's go home.' "
Metro's safety office developed six recommendations for changes after the incident.
One was that engineers should improve track circuit design, records show. Other recommendations said engineers should upgrade software used to monitor the circuits and formalize the monitoring process.
On the day of this summer's crash, documents show that the Tri-State Oversight Committee's database continued to list all six of the recommendations as "open." For each recommendation, the database shows that a corrective action plan was listed as still "to be developed."
Metro said that although it never told the oversight group about its actions, it had made the needed fixes and developed a software program to check for problems with the system. Officials said they used the software to look for circuit malfunctions once a week in the days after the 2005 incident. A year later, they dropped the frequency to once a month because they said they found no problems. Since the crash this summer, they have increased software checks to twice a day.
The NTSB concluded that Metro's crash avoidance system is inadequate and urged the addition of a real-time backup technology.
It is impossible to determine whether a more aggressive response to the 2005 incident could have averted this summer's crash. But federal officials have long considered Metro's slowness in closing out safety investigations to be a critical concern.
Records show the Tri-State Oversight Committee was concerned by Metro's lack of response to the 2005 incident even before the Fort Totten crash. In April of this year, after the near-collision on Capitol Hill, the committee drafted a letter to Metro's chief safety officer, asking her for the status of the six recommendations and noting that the committee had been "unable to verify the progress and implementation."
But the committee, which has no direct authority over the subway system and relies on Metro's willingness to share information, deleted any mention of the 2005 event from the final version of the letter. The change was made in the hope of expediting Metro's release of details about the March 2 incident, internal e-mail shows.'The Adrenaline Was Pumping'
Mitchell, 58, remembers the June 7, 2005, near-collision in granular detail.
During the evening rush, he said, he was pulling away from Foggy Bottom at up to 59 mph, carrying about 800 passengers in what operators call a "crush load." Unknown to Mitchell, just ahead in the tunnel below the Potomac, a train had stopped mid-river, waiting for the Rosslyn platform to clear.
Mitchell saw the train ahead and hit his brake at 5:55 p.m., records show. He could do little more than come to his feet, brace himself against the control panel and watch as his train stopped less than half a car length from the train ahead.
"The adrenaline was pumping, I'll tell you that," Mitchell said. "That is too damn close."
Moments later, he said, he heard the train operator behind him warn of an impending collision.
"That's when I became afraid," Mitchell said. "That's when I knew something was very wrong."
The entire incident lasted seven minutes, records show, and played out far from any exit or access point for rescue teams. No one was injured, and Metro gave Mitchell an award and a $1,000 bonus for his quick reaction.
When news broke about the crash this summer, he said, "It put me right back in that same place."
Mitchell attended the funeral of Jeanice McMillan, the train operator killed in this summer's crash. There, for the first time since his retirement this year, he saw the operator of the train that had come to a halt 20 feet behind him. They made small talk, Mitchell said, but made no mention of the four-year-old event that brought them both to pay their respects.
Staff writer James Hohmann contributed to this report.