Networked system tracks vibrations

PECO Energy installed a network-based system to monitor the condition of critical equipment at its Peach Bottom nuclear plant

Wade Mackey, David Hinrichs and Scott Danehower, PECO Energy

Graham Bleakley, Solartron Corp.

Much expensive power generation equipment is vulnerable because the equipment protection and condition monitoring systems are not adequate. Large generators and motors, particularly, benefit from condition-based maintenance to reduce downtime and enhance the possibility of extending plant life. The solution is often to graft improved condition monitoring onto existing equipment. Distributing intelligence around the sensor subsystems allows existing computer systems to carry on with their main task of controlling the equipment. Basing the improvements on international standards simplifies the upgrade and generally minimizes upheaval. Nowhere is this more evident than in nuclear power plants.

PECO Energy Co., formerly the Philadelphia Electric Co., is one of the largest investor-owned utilities in the northeastern United States, serving the state of Pennsylvania. The company operates seven plants, ranging from 160 MWe to 2,200 MWe. PECO Nuclear, one of the electricity generating divisions of PECO Energy, is focusing on reducing total operating costs at its two large nuclear installations at Limerick and Peach Bottom.

Recently, PECO concluded that the Peach Bottom turbine supervisory instrumentation (TSI) system and motor vibration monitoring systems needed to be enhanced. Several critical, continuously running pieces of plant equipment were lacking vibration monitoring, making it impossible to trend or diagnose developing problems. Even for those critical components with continuous vibration monitoring, the depth and quality of data did not allow accurate diagnosis. The monitoring equipment was also unable to capture data during machine transients, thus missing important changes in vibration levels during periods of worst-case stress. Shaft-rider probes, which were prone to wear, mechanically provided vibration signals. The vibration trip logic on the reactor feed pump turbine was based on a one-out-of-one signal from a shaft-rider probe, which provided many opportunities for inadvertent turbine trips. For example, in March 1996, prior to commissioning the new monitoring systems, a cleaner accidentally contacted a loose conduit. This agitated the vibration probe, causing a signal spike and subsequent trip.

The analysis system was also unable to correlate the various plant data, making it unable to relate vibration level changes to pressures, temperatures and flow levels. This correlation is essential in order to know whether increased vibration is due to higher turbine speeds or to loadings.

System constraints

At Peach Bottom the main turbine is inaccessible during plant operations, and individual components of the existing TSI could not be replaced without defeating the entire system. All bearing work on the main turbine generator had to be done within a two-week scheduled outage window, closely coordinated with other outage tasks in the same area. Despite the complex and interlocking actions (about 9,200 separate work order activities), Peach Bottom completely replaced the old TSI with a new Solartron 1051 in the two-week period.

The new system incorporates alarm and trip contacts for the main turbine/generator, and the reactor feed pumps and turbines. The system collects, stores, analyzes and displays vibration data for predictive maintenance on these units, and for the condensate and reactor recirculation pumps/motors. A backup machinery monitoring system monitors the main turbines and the reactor feed pump turbines, and provides alarm and trip functions independent of the computer network. This arrangement prevents computer or network failures from either initiating or preventing a main turbine or reactor feed pump turbine trip.

PECO management estimates the upgrade will save about (US)$60 million during the remaining 20 years of plant life. Savings will come primarily from advanced maintenance planning, which will reduce downtime and hence avoid lost power generation. Further, one-time savings have been identified as a benefit of the advanced system design and installation plan. For example, using the existing data highway instead of adding dedicated 4-20 mA dc loops for remote terminals saved more than (US)$50,000 in cabling and installation costs alone. Maintenance outage lengths also will be reduced. For example, when the steam turbines are restarted following maintenance, the individual turbine rotors tend to bow due to uneven heating. Before, engineers had to follow the turbine manufacturer`s worst-case recommendations, removing the bow using low-speed turning gear for four hours. Shaft runout readings from the vibration system can now verify that the bow is removed in less than an hour.

Most of the computerized data acquisition systems in the Peach Bottom plant are now integrated. The system already forms one of the world`s largest real-time vibration monitoring systems, and will soon be extended even further. It will then encompass more than 270 vibration analysis channels concurrently diagnosing 18 separate pieces of rotating equipment in a number of widespread plant locations. An essential part of the new design is to import and export data to other plant monitoring systems. These systems were already connected across a bridge to a token ring-based LAN and the corporate WAN, transmitting data throughout the company.


The new condition monitoring subsystem is grafted onto the existing Ethernet “backbone” of the plant`s data acquisition system (DAS). This links the various plant areas to a communications panel in the main control room, where bridge and server devices route the signals to the appropriate network segments (see figure). The subsystem continuously monitors and analyzes the plant across the existing data acquisition infrastructure (many of the new cables were installed in existing trays). It inputs vibration data to the existing plant computer and extracts related operational data.

The subsystem is based on two existing DEC Alpha workstations, which run DEC`s OSF/1 operating system and use UNIX-based application software. The workstations work in dual-redundancy “mirror” mode for hot backup. One workstation runs in primary mode–collecting, analyzing and archiving data–while the second workstation monitors the first, and takes over if there is a fault. The machines communicate across Ethernet using TCP/IP stacks, protected from the outside world by multiple fire walls.

Fiber-optic Ethernet links give immunity to electromagnetic interference in what is an inherently electrically noisy power generating plant. Ethernet also simplifies cable pulls, reduces power drain and ensures high reliability. For Peach Bottom, Ethernet`s other important feature was its open architecture. The whole upgrade is, in fact, a tribute to standardization, starting from Solartron`s 1051 on-line condition monitoring system. This uses VMEbus-based hardware, UNIX/OSF operating systems, OSI and TCP/IP networking, and X-Windows/MOTIF graphical interfacing. Seven of these 1051-based subsystems feed Ethernet data to the workstations. They provide extensive signal and data processing to reduce loading on the host.

The 1051 forms the primary condition monitoring component for two 1.1 GW generators. Each subsystem monitors 48 vibration inputs, with one synchronization input. Another 44 probes are field-installed as spares in locations not accessible during full power operations. Probes can then be changed without disassembling the bearings, thus minimizing the off-line time needed for making repairs.

One 1051 also monitors each of three reactor feed pump turbines, and one more monitors three condensate pumps. These condensate pumps originally had a common high-vibration alarm in the control room, which meant that if one of the pumps caused an alarm, this would “mask” the other two pumps. This alarm masking was solved by having separate high-vibration alarms for each pump.

A final 1051 monitors two of the recirculation pumps for shaft cracking. Throughout the nuclear power industry, there is a concern with possible shaft cracking on reactor coolant/recirculation pumps. The important parameters for predicting a shaft crack are the 1X and 2X vibration amplitude and phase readings. Adverse trends in any of these parameters may indicate the onset of a shaft crack. Users can establish “alarm envelopes” around the baseline readings for each parameter. An alarm window alerts the operator if any readings deviate from the preset envelope.

On-line monitoring

The 1051 provides up to 64 parallel dynamic analyzer channels and 120 static analog or digital data acquisition channels. It relates machinery vibration levels to the plant status and other static parameters like temperature, pressure and flow. It is this feature which integrates the condition monitoring system, providing valuable information during run-up, on-load and coast-down. The 1051 builds a machine life-cycle database of critical plant equipment to simplify fault diagnosis. This helps detect and identify potential equipment problems well in advance, again minimizing costly downtime.

A high-performance digital signal processor provides front-end processing, calculating the acceleration, velocity, displacement or dynamic pressure. It computes the complete set of key machinery analysis parameters, including harmonic analysis of up to 16 rotational orders with phase, peak, RMS, DC-gap and wide-band spectral information. The unit has advanced narrow-band analysis, and can maintain synchronous lock at ultra-low rotational speeds. A swept-sine analysis focuses on the harmonics required, while maintaining strict phase and amplitude accuracy. The processor validates and linearizes data, and converts it to engineering units before passing it to the host. Measurements are precise: the 1051 has built-in calibration signals, continuous transducer and wiring integrity checks and integrated board-level diagnostics.

Each 1051 has four auto-ranging A/D converters. These can be configured as four parallel synchronous/asynchronous channels or two simultaneous channel pairs. Each pair is typically used to measure signals from an x-y mounted transducer–for example, checking the vertical and horizontal vibration components to indicate shaft alignment. A great strength is the 1051`s ability to acquire vibration data in parallel with processing spectra; the system can also acquire simultaneous data from multiple rotating shafts. A dual-redundant tachometer supports duty/standby machine configurations (the backup speed transducer takes over if the primary fails).

Alerting operators

Plant operators are provided with a machine-state sensitive graphical interface which enables them to be directly involved during run-up, coast-down and on-load. An “alert” area displays any parameters that have deviated from pre-defined setpoints, and notifies the operator of any system errors. Coupled with advanced warning messages, these facilities enable plant operators to save time during variable speed operations–as much as 10 minutes in a 30-minute run-up.

The screen has a large display area for viewing incoming data as tables, bar charts, spectra and so on. Typically, these displays monitor both transient and steady-state conditions. However, the operator environment is strictly for viewing data and other important system parameters. No configuration changes can be made while in this mode.

The password-protected “analyst” environment provides many useful tools for vibration analysts to assess machine condition and diagnose problems, including user-configurable alert and alarm functions, data management and a wide range of analytical displays. Traditional orbit plots– frequently used during preload and to investigate shaft instability–can become cluttered rapidly. Instead, a “filtered” orbit plot using dc value, rotational order or a combination of orders, provides an uncluttered display and allows analysts to focus clearly on specific machine problems. Waterfall and cascade plots of spectral data make it easy to monitor machines during a speed transient. Data can also be shown on bar graphs, Bode plots, multi-plots (showing dissimilar data, like vibration and temperature, on the same plot), Nyquist plots, polar plots, shaft orbits, shaft displacement, spectra, trend plots and time waveforms. There is also an event log and a notebook.

The analyst can select live or historical data, using one large display window or four smaller windows showing four different plots simultaneously. This is especially powerful when comparing the behavior of similar pieces of equipment. The display can show relationships between seemingly unrelated vibration and static plant parameters, such as highlighting interactions between turbines and auxiliary equipment. Comprehensive trending of recent or long-term history enables several linked parameters to be compared quickly to reveal impending faults at an early stage. For instance, at Peach Bottom there was a problem with an outboard pump bearing. The smaller windows showed the outboard pump shaft orbits from three similar pumps, highlighting the differences.

The data recording and replay functions provide rapid access to stored data, while long term data can be archived for lifetime care. An on-line “study process” gathers the real-time and historic data into a workbook of plots and observations for future reference. END

Author bios

Wade Mackey is a design engineer in the site engineering group at Peach Bottom Atomic Power Station. A graduate in engineering from Widener University, Mackey has 18 years experience in the power plant field, including six years in instrumentation and control.

David Hinrichs is system manager and network administrator for the real-time plant monitoring systems at Peach Bottom. Hinrichs` eight years with computers at Peach Bottom was preceded by eight years in power plant startup at Bechtel Power.

Scott Danehower is a component engineer in the site engineering group at Peach Bottom. He is a graduate in mechanical engineering from Widener University and a Pennsylvania-registered Professional Engineer.

Graham Bleakley joined Solartron in 1979 having gained a Higher National Diploma in applied physics from Thurso Technical College (Scotland). A condition monitoring specialist in the data acquisition group at Solartron, he currently has worldwide responsibility for exploring future requirements in advanced turbine monitoring systems.

Click here to enlarge image

Solartron 1051 vibration monitoring system undergoing final checks at PECO`s Peach Bottom nuclear station.

Click here to enlarge image

Click here to enlarge image

PECO`s nuclear power generation facility at Peach Bottom, Penn., USA