News from the pit

9 June 2008

HAUG Gas Compressor Used in the Evaporative System


The teams commissioning the pixel detector last month had to know a thing or two about keeping their cool, not least those involved with testing the cooling system, which gave them not one but two riddles to contend with. Steve McMahon, who is responsible for the Inner Detector environment including cooling, spoke to the e-News about the difficulties they’re currently facing.

“There are basically two separate issues,” according to Steve. The first of these concerns the temperature control system linked with the pixel cooling loops which carry coolant around ATLAS; the second – rather more serious – problem has arisen from an unexpected failure in an area of the external cooling plant, which delivers coolant to the detector.

Testing began on the cooling loops on the morning of April 25th. Over the following few days, 87 of the 88 pixel loops were commissioned successfully, but with some unexpected temperature fluctuations evident in a number of circuits.

When all is working properly, cold liquid exits the detector at -25˚C. Just before it passes into the cavern from within the humidity-controlled environment of the Inner Detector (ID), it must be heated to prevent the pipes in the cavern – where the relative humidity is below the dew point – from freezing in an instant.

The heater control system works on a feedback loop which constantly monitors the temperature of the exiting liquid, and adjusts the amount of heat being added at the exit accordingly. During commissioning, however, some circuits showed large fluctuations in temperature. In a bid to factor out the automatic temperature control element, the commissioners “tricked” the system by turning on the detector, which acted as a surrogate heat load. But, says Steve, “Even when we factored the control process out, we still had these oscillations. We were investigating those when the plant failure happened.”

The failure occurred on Thursday May 1st, when three out of six coolant compressors “failed in a catastrophic way”. When the company who supplied the compressors, Haug, arrived the following morning and opened up the damaged units, there was evidence of crackling in the casing, and they diagnosed the compressors as having burnt out.

“The compressors should be driven by a magnetic coupling,” explains Steve. “They failed because they slipped – the motor started, but the magnetic drive wasn’t transmitted to the crank that does the compression. We’re fairly convinced that the problem started the day before, and that the three that were slipping were doing so for something short of 24 hours.”

“We’ve also found things inside that we don’t understand,” he adds. “Inside the one which is the most severely affected, we found a water leak coming in from the coolant that circulates around the compressor. It’s failed in a bad way, but we don’t yet know how this part relates to what happened to us.”

Haug gave the other three units in the cooling plant a clean bill of health when they came on May 2nd, and took the damaged ones away for closer inspection. The ID teams (pixel and SCT), together with colleagues from the TSCV team, have spent several weeks investigating the problem alongside the company, and now believe that the failure is limited to the plant. This is welcome news, as it means that the detector itself is not compromised.

A clean-up operation is now underway to purge the system of the thick black powder which filled it during the plant failure, and new C3F8 liquid freon has had to be ordered. In light of similar problems occurring outside of CERN, Haug began adding extra sensors to the units they produce as of Christmas 2007. The three refurbished compressors will soon be at ATLAS, and will be fitted with this new technology, meaning the ID teams and cooling plant operators can keep a much closer eye on what is going on within the units. This way, catastrophic consequences will be avoided in the event of any similar slipping faults occurring in the future.

As for the instabilities in the re-heating of the cooling loops, progress is being made here too. The Inner Detector is now completely closed, but before it was, the pixel team added some additional sensory instrumentation. “This means that, when we start up again, we’ll have a much better handle on what went wrong,” says Steve, “and then we can plan how we’re going to tackle it.”


Ceri Perkins

ATLAS e-News