01 June 2003
An outage that needn't have been
By Brad True
Not preventive maintenance or trend monitoring software, this technology provides equipment-specific, condition-based early warning.
Boilers, turbines, generators, and auxiliary systems used in power generation are high-value assets that cause expensive problems when they fail.
End users in virtually every industry are adopting asset management as a strategy to improve process efficiency and return on assets.
One large power producer in the U.S. began a project to investigate information technology solutions, especially predictive maintenance and equipment condition monitoring solutions, to provide early warning of impending equipment failures.
Unlike traditional preventive maintenance or trend monitoring software, these solutions provide equipment-specific, condition-based early warning.
By analyzing data from a previous failure using a multivariable strategy and by looking at logically related sensors, the equipment condition monitoring solution reported equipment failures earlier than traditional methods.
This early warning would have given the plant operators time to significantly mitigate the eventual consequences.
Unlike preventative maintenance practices, which recommend maintenance based on failure statistics for a class of equipment problems over time, an entirely new class of solutions, called predictive maintenance and equipment condition monitoring, has emerged to provide equipment-specific, condition-based early warning.
Quite simply, this tool provides advanced, machine-specific warning of deteriorating conditions leading to failure and/or poor performance.
GENERATING UNIT SUFFERED
In the power generation industry, this system provides early warning of failure for assets such as combustion turbines, steam turbines, boiler feedwater pumps, and cooling water pumps.
By acting upon specific, condition-based warnings, companies greatly reduce catastrophic failure frequency, mitigate the impact of deteriorating equipment conditions, increase asset availability, decrease maintenance expenditures, and recapture the hidden costs of downtime.
A major Midwest fossil power plant operator investigated these solutions to deliver early warning of impending equipment problems. After a thorough review of software products, the company identified and tested the SmartSignal eCM solution.
The company provided historical process data from a steam turbine–driven boiler feedwater pump at a 600-megawatt nominal base load coal-fired power plant that had experienced an unplanned failure.
The test required the solution to demonstrate that the software could not only identify the problematic component, but also detect the signs of incipient failure earlier than traditional plant monitoring technology.
The power plant experienced an unplanned unit derate due to significant damage to the first stage wheel of the turbine-driven boiler feedwater pump.
The power generating unit suffered not only a complete loss of the turbine-driven pump, but also a unit derate because the motor-driven backup pump was not designed to replace the full capability of the failed equipment.
Technicians analyzed a data snapshot that included the six weeks before the steam turbine failure. The analysis consisted of data-driven empirical models that captured specific elements of the turbine-driven pump's operation.
The model could provide early failure warning based on the assumption that the turbine-driven pump operated normally during the first twenty-one days of the forty-two-day data set.
The pump model provided estimates of the vibration, pressure, temperature, speed, and flow associated with the pump. Second, the turbine model estimated the pressures, temperatures, flows, vibration, and bearing temperature associated with the turbine.
A third model grouped process variables that indicated turbine and pump power into a single model to detect efficiency losses.
DISCOVER VIBRATION ANOMALY
Plant personnel applied the models to the last three weeks of data to detect early warning of failure. Neither the pump model nor the pump-turbine efficiency model revealed signs of failure. The turbine model, however, provided very interesting results.
The software illustrates the actual and estimated vibration signals associated with the turbine. The sensors include inboard and outboard bearing radial shaft vibration, axial shaft vibration, and inboard and outboard bearing rotor vibration.
The graphs show that both rotor vibration signals alerted briefly following pump startup and then alerted continuously following the pump startup two days later. The turbine failed catastrophically four days after that when the vibration jumped again and tripped the high-vibration interlock shutdown.
The vibration alerts are notable because they flag process variations, which are within the turbine protection system high- and low-vibration threshold limits, and because they came six days before the catastrophic failure.
However, a reliability engineer with sufficient analytic capability could study the turbine rotor vibration phase diagrams to discover this vibration anomaly.
A separate piece of evidence leverages the turbine process variables. The turbine first stage pressure alerted continuously for four days prior to failure.
The lower than expected turbine first stage pressure suggested a loss of power extraction capability from the turbine indicative of the actual failure mode—the partial loss of the first row of turbine blades.
The process data deviations, coupled with the vibration deviations, provide a compelling case that the turbine deserved maintenance attention so the plant staff could minimize the consequential damage to the turbine before the catastrophic failure finally occurred.
Significantly, the software alerted continuously regarding turbine rotor vibration and first stage turbine pressure. The alerts were not only early warning of a catastrophic failure, but they also identified the faulty component at least four days before the actual failure.
After the steam turbine shut down due to high vibrations, the maintenance team disassembled the steam turbine and discovered damaged turbine blades over the entire first row. The ability of this tool to analyze the critical data and isolate the specific failure mode could have given the power generator early warning of the impending failure in the turbine's first stage wheel.
By taking action, the power plant could have prevented the catastrophic turbine failure. Early action reduces maintenance expense by limiting the extent of the turbine damage and minimizing the duration of the forced outage. IT
Behind the byline
Brad True is the general manager—energy, power, and process at SmartSignal Corporation in Lisle, Ill.