Condition monitoring compares, contrasts offshore equipment performance

Darren Witherwick
MDC Technology Ltd.
Teesside, U.K.

The application of model-based condition monitoring results in more effective maintenance scheduling by providing improved data for process operator and engineer analysis.

These systems monitor key process equipment, such as gas turbines, pumps, and compressors. One such system is installed on A/S Norske Shell's operated Draugen production platform in the Norwegian sector of the North Sea and will be discussed in this article.

A condition-monitoring system uses process models of operating equipment to predict key indicators. The Draugen system also uses a data historian/management information layer to provide for data analysis such as trending and bar charts.

The technology described and the general principles of model-based condition and performance monitoring are applicable to many industries. But these systems are particularly useful in offshore applications, which have inherent problems in equipment replacement and maintenance.

Model-based condition monitoring

The main driving force for model-based condition monitoring of machinery is to detect and diagnose technical problems at an early stage, thus allowing for more-effective planning of maintenance strategies.

Furthermore, it is desirable for technical maintenance intervals (inspections and overhauls, etc.) to be lengthened by using condition-based maintenance planning.

All process equipment performance degrades with time. Not only does performance degrade between overhauls but also with each successive overhaul, the performance gains are less (Fig. 1 [51,228 bytes]).

Applying a model-based conditionmonitoring improves decision-making based on derived performance and economic data, enabling guesswork to be removed from maintenance scheduling.

Process models

Process models are at the core of any performance-monitoring system. These models are mathematical representations of each of the equipment units monitored by the system. They are based on standard engineering principles and use both design and actual performance data to accurately simulate equipment operation.

It is imperative that the process models continuously and accurately represent the actual equipment units, so that performance data are generated from a validated model that defines the current plant behavior.

All process models are designed with intuitive engineering tuning parameters. These are used to tune the models on-line against actual unit performance. This accounts for in-service performance deterioration.

The tuning process, model updating, is essential to the efficient operation of any modeling system and is performed on a continuous cyclic basis.

Model updating is achieved by employing a least-squares algorithm to determine a "best" fit to plant data by adjusting the identified key model parameters. For each parameter evaluation, multiple sets of time synchronous plant readings are used. This avoids biasing the models based upon process noise.

Performance degradation

The key concept of model-based condition monitoring is the ability to compare and contrast process performance in the clean or design condition (often referred to as the day-zero condition) against performance in the current degraded condition.

This is achieved by running in parallel two instances of the process models. One instance simulates operation in the clean state, and the other simulates operation in the degraded state.

By applying suitable costing criteria to plant performance, for example differential fuel costs to generate equivalent power, the costs of performance degradation can be quantified.

These degradation costs can then be contrasted with the cost of equipment outages/cleaning cycles to provide the necessary justification for cleaning.

The clean and degraded conditions are simulated using the same basic engineering models with different tuning parameters. The day-zero condition is simulated by tuning the model performance to match design data.

The degraded condition is a result of applying the current model tuning parameters.

Draugen system

The Draugen platform has an MDC Technology condition monitoring system (CMS). The CMS is an advisory computer system for use by both plant operators and engineers and is based on MDC Technology's "Appreciate Power" product.

The system monitors key rotating machinery and replaces a Boyce DATM4 system.

On the Draugen platform, the critical machinery monitored include:

• Three GT35 (ABB) gas turbines driving ABB generators
• Two EGT gas turbines driving water-injection pumps
• Four Demag gas compressors
• A number of pumps, including crude oil and seawater pumps.

An Advant Station 500 IMS (information management system) server is the user interface to the Draugen platform ABB DCS (distributive control system), and the data source for the CMS. The system supports the AdvaInform relational data base, which is based on the Oracle relational data base management system (Rdbms).

The IMS operates on an HP-Unix platform. The CMS system receives process data from the platform once per hour via the IMS.

The CMS uses the MDC Technology RTO+ plant modeling suite and the OSI Software data archiving tool PI (plant information).

The CMS system supports the following features:

• Real-time interface to the IMS system
• Data acquisition and archiving
• Data trending
• Graphical analysis
• Numerical reports (both predefined and user configurable)
• Process modeling and performance monitoring functions
• Simulation tool for user defined "what-if" style operations.

The CMS system operates as a client/server architecture. The server hosts the data acquisition and archiving tools as well as the process modeling and performance monitoring functions.

User access to the system data and data analysis tools is via client PCs. The server and PCs operate under the Microsoft Windows NT environment.

The CMS system has a number of constituent parts, as follows:

• Data historian
• Man/machine interface
• Performance monitoring software
• Data communications software
• Real-time executive.

Data historian

Data acquisition and archiving facilities are provided by OSI Software's PI data archive system. The CMS data base is sized to maintain both current values and historical data for up to 1,500 data points over the expected working life of the platform.

These include raw process readings (from the IMS), manipulated data such as ISO-normalized values, and derived performance-monitoring parameters.

Man/machine interface

The PC-based user interface to the CMS is hosted by the OSI Software PI-ProcessBook.

PI-ProcessBook is a PC tool for displaying plant information stored in the PI data archive and comprises a collection of information displays grouped together like the pages of a book.

The displays can show a variety of elements such as schematic representations of process equipment, plots of readings, and trends of tags.

Trend facilities are provided as standard in the PI-ProcessBook tool.

PC tools such as Excel and Olectra Chart are used to provide report generation and graphical analysis facilities. All numerical and graphical reports are embedded within Process Book displays using OLE capabilities of PI-Process Book.

The Draugen CMS interface enables the user to:

• View mimics depicting overviews of the major process areas on the Draugen platform, for example the gas-turbine power systems and the gas-compressor system.
• View mimics of each of the equipment units displaying all associated tags, both raw and derived from the model. Tags on the displays are dynamically linked to the PI data archive. The equipment mimics also indicate alarm states.
• Trend data in both real-time and against equipment running hours. Real-time is the default time basis.
• Trend data against design points at reference load conditions.
• Trend extrapolated data to infer when alarm limits will be violated.
• View graphical reports and bar charts, both preconfigured and user defined.
• View numerical reports (both preconfigured and user defined).
• Manually redefine the gas compositions for each of the four gas-compressor stages when new gas samples are provided.
• Pose and trigger "what-if" style simulation runs of key process units using dedicated displays accessible from the equipment mimics.

Performance monitoring software

The plant modeling and performance monitoring functions are provided by RTO+, the process simulation element of Appreciate Power.

These functions operate on a continuous basis, running at predefined frequencies on the server. The results of these functions are archived in the data historian.

The performance-monitoring software provides models of each of the key process units.

The Draugen gas-turbine units are two-stage machines consisting of a gas-generator unit combined with a free-power turbine section. They are modeled comprehensively in a two-stage manner.

Each section of the gas-generator unit (compressor, combustor, and turbine) is modeled in isolation. The additional free-power turbine section is treated as a separate expansion stage.

Design data for each of the stages is used to relate current performance to design performance.

The compressor models are based on design characteristic curves of compressor performance against the volumetric suction flow. Performance curves of polytropic head-vs.-suction flow and polytropic efficiency-vs.-suction flow are employed to define the pressure ratio across the compression stage for a defined suction condition.

Power consumption is calculated based on the isentropic efficiency of the machine.

The pump models are based on characteristic curves of pump performance against the volumetric flow rate through the pump. Performance curves of pump head-vs.-suction flow and pump efficiency-vs.-suction flow are employed to define the pump performance.

Discharge pressure is determined from the application of Bernoulli's equation. Frictional head losses, and suction and discharge point elevations are included in the calculations.

Updates of key model parameters for each process unit, to account for equipment degradation, ensure that the process model continually reflects actual plant performance. These include:

• Efficiency factors applied to each stage of the gas-turbine model, including the power-turbine section.
• Offsets to translate the compressor performance curves.
• Offsets to translate the pump performance curves.

Key performance indicators for the plant are generated on current plant operating conditions, ambient conditions, and the current model parameter set. The indicators infer equipment operation efficiency, degradation, fouling effects, deviations from expected performance, and operation at standard conditions.

The information produced enhances that available from plant instrumentation and provides indications of performance where plant instrumentation is not available.

Communications software

There are two data interfaces in the CMS application, as follows:

1. Data interface between the IMS and the data historian
2. Data interface between the data historian and the RTO+ layer.

Data communications between all components are handled by standard interfaces.

Data transfers between the components of Appreciate Power take place via MDC Technology's real-time data interface, the data exchange server (DXS+). This incorporates a PI API driver to allow communication between the PI data base and the modeling system, RTO+.

Data reads from the PI data base are controlled by DXS+ and are performed at a fixed cyclic scan rate. This ensures synchronization of the data to be used in the performance monitoring calculations.

The data exchange server also provides first-line data conditioning and validation. The features available include:

• Identification and elimination of obviously bad measurements, such as out of range values.
• Value substitution for bad or suspect readings. A number of alternative values can be selected for transmitting to the modeling software, such as the last good value, an hourly average, or a default value.
• Scaling of plant data. This allows plant data to be converted to consistent engineering units set at the interface layer.
• Filtering of plant data to remove noise and instabilities from the readings.

Real-time executive

The real-time executive is the controlling function that coordinates the running and interaction of the RTO+ processes.

The various processes operate in an asynchronous manner, exchanging data with each other via the communications interface, as shown in Fig 2 [61,905 bytes].

The Author