OPEN predictorTM
INTRODUCTION TO THE OPENpredictor CONDITION AND PERFORMANCE MONITORING
SYSTEMS
Copyright2005. All rights, title and interest in and the Software, Hardware and Services detailed in this document and all copyrights, patents, trademarks, service marks or other intellectual property or proprietary rights relating thereto belong exclusively to ROVSING Dynamics A/S
Marielundvej 41 · DK-2730 Herlev · Denmark Tel: +45 46 90 72 00 · Fax: +45 44 84 60 40 · VAT No. 20 05 24 73 e-mail:
[email protected] · www.rovsing-dynamics.com
CBS_OP-Intro_Jan05.04_03
TM
OPENpredictor™ CONDITION AND PERFORMANCE MONITORING SYSTEMS
Automatic fault diagnosis and health prediction
FAULTS ON ROTATING COMPONENTS Rotor/shaft: Unbalance Bent rotor Eccentricity Oil whirl Steam whirl Rubbing in bearings & seals
To assess the mechanical and functional health of plant assets, the widest range of potential machinery problems shall automatically be identified. During different transient and stationary operational states of the machines, the development of specific faults shall be identified. As the development of faults is related to machine operation, the measured data has to be classified to transient and operational states in order to achieve meaningful data comparison, automatic fault diagnosis (AutoDiagnosis™) and health prediction.
Blades & impellers: Rubbing Cavitation surge Stall
In the following overview, several machinery components, potential problems and their early diagnosis are presented. The OPENpredictor™ system can however be configured to identify a wide range of other machine specific problems when more detailed machinery information is provided.
Casing: Electrical excitation Misalignment Thermal uneven expansion Blocked bearing movement
Operation via mimics Operation via the mimics is also straightforward. Machine pictograms or photos are used to present sensor locations. Operator data is presented below the defined sensors. This data is automatically updated when new values have been identified. The color of the data represents the alarm status. By clicking on a sensor a graphical plot is started showing the trend data.
Operation via browser For the vibration expert or process engineers the browser is an effective way to retrieve any data from the database. Dedicated filtering functions allow selective display of data, which eases trouble-shooting. The experts can pre-configure preferred data displays in order to ease future data retrieval. The window contains four areas below the title bar: • Menus • Toolbars • Current item/measurement selected • Left field: Item/measurement selector • Right field: Item/measurement data
Combustor: Resonance Flame instability Flame distribution
Rolling element bearings: Outer race defects Inner race defects Cage defects Lubrication deficiency
Reporting Report generator OPENpredictor™ incorporates a report generator to create dedicated reports. These reports can be stored under user-defined names for easy retrieval. Apart from these reports any graphical plot can be exported in HTML format for easy transfer via an e-mail system.
Journal bearings: Shaft lift fault Wear
Machine components The overall health of a machine depends on the condition of all components critical for the machine operation and all dynamic forces acting on the components. The components are subdivided into rotating and stationary items such as: Mechanical and functional health assessment ROTATING COMPONENTS Rotor/Shaft Gear wheels Blades Couplings
FAULTS ON STATIONARY COMPONENTS Foundation: Looseness
Thrust bearings: Wear Couplings: Locking Gear wheels: Wear Back lash Tooth defects Pitting
STATIONARY COMPONENTS Foundation Casing Seals Bearings Combustor
Example of Machine Mimic with real time measurement values and links to other mimics and graphical plots.
Shift reports Shift reports will be printed out automatically at the specified time, after they have been defined (once). They will contain the requested operation information for the shift group in the control room, stating where special attention is needed.
It is possible to create different levels of mimics, which are actively linked together via "link buttons". This allows easy navigation between machines located in different plant sections without having detailed knowledge about machine names, tag identifiers and sensor locations.
Management reports Management reports have to be requested manually. These reports typically will contain maintenance information on identified machinery faults, fault predictions and recommendations. For the experts detailed data can be included to justify the maintenance activities.
For specific machinery design, additional potential faults can be specified.
When no design and/or installation problems exist, and then the typical sources of machine health deterioration are component fatigue, wear, erosion and external factors. Machine mechanical health monitoring provides information to optimize machine availability, while machine functional health assessment provides information with regards to machine performance (e.g. component efficiency).
Monitoring of e.g. gas turbine performance degradation Gas turbine performance degrades over time due to recoverable and non-recoverable mechanical changes. The result is a reduction of the power output and an increase of the heat rate (a decrease of thermal efficiency). Typical recoverable losses are fouling of blades (mostly compressor blades) and air filters. These losses can be recovered by operational cleaning procedures or by external maintenance.
Monitoring problems that influence machine availability The following list is an overview of the most common faults to be monitored in order to assess the mechanical machine health:
Typical non-recoverable losses are blade erosion or deviations in tip clearance and seals. These losses
2
The OPENpredictorTM browser provides hierarchical machine/ components/measurement overview and configuration details. From the browser, graphical representations of measurements can conveniently be invoked.
11
Fault diagnostic messages shall be acknowledged by the user and can simply be printed or routed to the e-mail system.
function minimizes machine operation risk, as any important identified change in machine behavior will give a warning. When OPENpredictor™ has identified an "alert" or "alarm", a "warning lamp" starts flashing, and by clicking the flashing "warning lamp" an "alert list" will show detailed information. One more mouse click reveals the information either as a graphical plot or as an AutoDiagnosis™ message. Accessing data is extremely simple using well-known icons and menu-bars.
Fault Diagnostic Message.
From the fault diagnostic message, the user can access a prediction plot (see example below), forecasting how the fault will develop over time. Operation via warning system This is a very efficient way to retrieve information about identified changes in the machine behavior, for which no AutoDiagnosis™ can be provided yet. This
The Prediction curve presents graphically the specific fault Prediction, together with the confidence range.
can only be recovered by replacement of internal gas turbine components.
(residual lifetime of concerned components). The following illustrates some of the signatures used in OPENpredictor™ together with their main fault identification capabilities.
To monitor the deviation from a new gas turbine performance, is valuable to schedule compressor cleaning. The cleaning and filter exchange date can be forecasted and performed on the most economical date. Additionally, the long-term retrofitting can be economically optimized. The performance monitoring system uses existing process measurements such as temperatures, pressures, fuel specification, ambient conditions, etc. The system provides actual and corrected (corrected to reference ambient) estimates on: • Power output • Heat rate, thermal efficiency • Compressor efficiency • Turbine efficiency • Fouling indices for compressor and filter Process problems that influence machine availability Potential dangerous process circumstances can cause machine damage such as: • Cavitation • Surge • Stall • Combustion pulsation
The Constant Percentage Bandwidth Signature (CPB) gives a general health overview of the majority of mechanical fault characteristics, such as, unbalance, misalignment, foundation looseness. It is however, specifically sensitive in identifying faults such as rolling element bearing faults, cavitation, combustor resonance and gas leaks.
When these problems occur, the operation shall be changed in order to avoid strong dynamic forces to blades, impellers and seals. The result is reduced operational risk and longer service life.
OPENpredictor™ machine health assessment methodology Early fault identification OPENpredictor™ provides a unique library of signatures dedicated to detect and identify most mechanical problems, which may be encountered for common rotating and reciprocating machinery. The dedicated signatures check for different fault characteristics and provide selective information regarding fault symptoms, fault development as well as fault locations. Information provided by the different signatures form the basis for automatic fault diagnosis (AutoDiagnosis™) and prediction of fault development
10
The Autospectrum Signature (FFT) is used to identify faults which can only be diagnosed with a high frequency resolution, such as electrical excitation, distinction between electrical and mechanical imbalance, blade passing excitation and system resonance.
3
provides a high sensitivity for early fault identification and is used for AutoDiagnosis™ and prediction.
Information retrieval The OPENpredictor™ system has been designed for unmanned operation. As machines are typically operating reliably over a long period of time, it is essential that a PMIS system automatically provide warnings, when actions need to be defined.
OPENpredictor™ server
The Cepstrum Signature (CEP) provides a good selectivity to identify gear wheel faults in complex gearboxes and any other faults, which result in a signal with modulation character.
The Selective Envelope Detection Signature (SED) is specifically sensitive to identify faults where symptoms are of repetitive impulse character, such as rolling element bearing faults and cavitation. The AutoDiagnosisTM function uses the repetition rate of the fault to conclude the origin of the fault.
The OPENpredictor™ server receives the data measured and processed by the SPU's. The server stores the data in the internal RAM memory, from which it is transferred into the relational database. In this way large bursts of data will not result in data loss. If data variations are less than a user-defined percentage from previously stored values, then data is automatically removed. This assures that the database is not filled with data, which does not provide any valuable information. Alarm information and fault diagnostic results are also stored in the database.
There • Via • Via • Via • Via
Operation via AutoDiagnosis™ messages OPENpredictor™ automatically interprets changes in machine behavior and will issue AutoDiagnosis™ messages when machinery faults have been positively identified.
The OPENpredictor™ server also runs the AutoDiagnosis™ and health prediction programs, data display and reporting programs as well as the remote access program.
A distinction is made between instantaneous and gradually developing faults. An instantaneous fault does not have a prediction e.g. a shaft rub or cavitation, as it occurs during a certain period as a coincidence of operation circumstances.
Man Machine Interface The user The man machine interface has been developed to meet different user requirements. Users are typically operators in the control room, maintenance engineers, process engineers and system managers. As these users have different responsibilities and demands, the system can be configured to meet individual information requirements.
The Transient Signature (TRT) is used to automatically identify changes in the run-up or coast-down behavior of a machine. Changed critical frequencies and decreased damping are automatically diagnosed.
Vector Analysis is used extensively in the AutoDiagnosis models. OPENpredictor™ provides a unique implementation of Order Tracking Analysis (OTA) that provides accurate vector values, both in steady state and transient conditions. A vector consists of both an amplitude and phase value, referencing to the position of the shaft at the tacho probe. The unique algorithm provides the vector values for four harmonic components and the sub-harmonic representing shaft instability.
Wear phenomena in bearings and gears however typically have a long development path, for which a prediction will be presented. The AutoDiagnosis™ message pops up automatically and provides information about the identified fault, the component involved, the fault strength and if configured by the plant a recommendation of the action to be performed. If a prediction is available then the fault strength can be plotted as a trend together with its expected future development to assist in planning of the corrective task.
OPENpredictor™ incorporates three different warning systems, one each for operation, maintenance and process optimization. System operation is similar for each user but the type of information presented can be adapted to the plant and user requirements.
Alarm Bar
4
exists four different ways to operate the system: AutoDiagnosis™ messages the warning systems the mimics the browser
9
SIGNATURE PROCESSING SYSTEM
The Circular Analysis Plot (CA) is used to monitor changes in symmetry for exhaust gas temperatures, thrust bearing pressure and temperature.
RO2000 Signature Processing Unit (SPU)
The advanced library of signatures provided by the OPENpredictor™ system supported by the very latest technology and performance in signal processing as implemented on the OPENpredictor™ signal processing module, paves the way for detection and diagnosis for a wide variety of all mechanical machine faults. Only a limited number of sensors are required, of which most may already be available on the machines to be monitored. Signals from field sensors are routed to the OPENpredictor™ signature processing unit (SPU), which consists of up to ten signal-processing modules, each handling eight sensor inputs. Processing in the SPU is performed by the individual signal processing modules. Parameters and signatures (including alarms) are transferred to the OPENpredictor™ server for AutoDiagnosis™ processing, prediction, reporting and display. All results are stored in the OPENpredictor™ database for retrieval later, as required.
Signature Processing Unit (SPU) The SPU will be configured from the OPENpredictor™ server. The SPU performs data acquisition, signal conditioning, classification and signature calculation functions. It also performs continuous alarm checks on scalar and vector data as well as on the calculated signatures.
The Shaft Centre Line Plot (SCL) shows the location of the shaft centre in the bearing clearance. The acceptance location domain is indicated by the yellow/red alarm limits.
Sensors Existing sensors Existing sensors will be used as much as possible. The signal output of existing vibration sensors can be duplicated to the SPU inputs using galvanic isolators or direct-buffered sensor outputs of already installed vibration-monitoring panels. This reduces engineering and installation activities to a minimum. Modulation spectrogram as basis for SMD-function.
Load and Reactive load signals will be directly measured by the SPU for the major operation Classification. RO2020-1 Signal Processing Module (SPM)
Other process signals, e.g. for gas turbine efficiency monitoring will be duplicated using isolators in the marshaling racks or a serial link from the control system.
Scalar dynamic data On-line scalar data are overall vibration values, position values, axial positions and machine speeds.
Additional sensors The machines will be equipped with additional dynamic sensors as needed, in order to calculate all the signatures needed for complete fault detection and diagnosis.
Vector data The SPU processes the raw vibration signals to calculate vector values. These vector values improve the AutoDiagnosis™ functions and provide the basic input to the balancing program (option).
SMD function showing synchronous modulation strength as function of time.
Signature calculation and analysis The signal processing modules acquire data and perform the signature calculations. Signature analysis
RO1000 Accelerometer (OEM product).
8
5
Classification of signature processing Machine behavior (and consequently measurement performed on the machine) depends on how a machine operates. Therefore, calculated signatures will also vary with machine operation. Machine health assessment has to be performed under comparable operating conditions and fault symptoms shall be assessed only within a given machine state. That is accomplished by OPENpredictor™ and this concept is referred to as "Classification of Signature Processing, Fault Diagnosis and Prediction".
Left: Fault diagnosis, severity and recommendation are provided in clear text. User recommendations can freely be added in order to assist the operators in their daily task. Clicking with the mouse on Explorer Graph reveals the forecast curve. This fault description with severity estimation and recommendation can then be directed to a maintenance management or reporting system.
It is important to note that a number of faults will only reveal themselves in some or even in one machine state and that some faults will only reveal themselves during transient states, such as run-up/coast-down. In conclusion, Classification is essential to improve identification sensitivity, to create realistic prediction and to avoid false warnings.
Predictions are available in the defined machine states with the benefit of giving critical dates for alternative operating conditions. For faults where a combination of parameters is required for diagnosis, the AutoDiagnosis™ function provides fault symptom prediction.
Typical start-up behaviour of a gas turbine Engineers' tools Detailed manual analysis is possible using the signatures plots as well as the "manual analysis tools" such as Orbit, 3-D, Multiple Trend Analysis, Scatter and Nyquist analysis, etc. Statistical analysis functions are available for easy data classification and investigation of relationships between parameters.
Level
The AutoDiagnosis™ fault libraries are expandable to new experience and new machine designs.
1
2
3
Steady States
4
5
Typical Transient and Steady States:
Transient States Gradient States Calibration
The time a machine is not producing is defined as
. Production states are defined as . TRANSIENT STATES State 1 Calibration State 2 Slow roll State 3 Ventilation, temperature setting State 4 Run-up (cold and/or hot) State 5 Over speed testing State 6 Synchronization Etc.
6
7 1. 2. 3. 4. 5. 6. 7. 8. 9.
8
9
Time
Test phase Slow roll Ventilation Run-up, cold, warm, hot Over speed testing Syncronisation to mains Low load Medium load .............
Prediction of fault development Prediction of fault development is performed on selective areas in signatures, for which the level relates to specific machinery problems. The system automatically forecasts the date that a fault symptom will reach the defined alarm level. As all data is classified to comparable process circumstances, the prediction provides a reliable estimate.
Extensive data filter and sorting facilities allow fast and easy access to historic data and a wide scope of graphic tools provide presentation in any appearance domain for easy assessment and troubleshooting.
Automatic fault diagnosis (AutoDiagnosis™) OPENpredictor™ provides automatic fault diagnosis for faults in critical machine components and potential faults. The foundation of AutoDiagnosis™ is the synergy between classification, normalization and fault selective signatures. The result is early fault detection and accurate diagnosis with diagnostic messages presented to users in clear text. Locally defined recommendations for action, together with prediction of fault development as well as date for required inspection/maintenance ease the tasks for the users.
OPERATIONAL STATES State 7 Low load, inductive operation State 8 Medium load, inductive operation State 9 High load, inductive operation State 7b Low load, capacitive operation
Fault prediction provides the user with clear information required for maintenance planning or operation improvements.
State 8b Medium load, capacity operation State 9b High load, capacity operation Etc.
6
Any measured signature can be plotted in a 3-D presentation as function of time, RPM or any other process parameter to show dependencies. Applying data filters removes data, which is temporarily not relevant.
The 3-D representation may also be viewed from a birds perspective in the so-called Campbell plot. Here, the Frequency-Time dependent amplitude is shown in colour code.
7
Classification of signature processing Machine behavior (and consequently measurement performed on the machine) depends on how a machine operates. Therefore, calculated signatures will also vary with machine operation. Machine health assessment has to be performed under comparable operating conditions and fault symptoms shall be assessed only within a given machine state. That is accomplished by OPENpredictor™ and this concept is referred to as "Classification of Signature Processing, Fault Diagnosis and Prediction".
Left: Fault diagnosis, severity and recommendation are provided in clear text. User recommendations can freely be added in order to assist the operators in their daily task. Clicking with the mouse on Explorer Graph reveals the forecast curve. This fault description with severity estimation and recommendation can then be directed to a maintenance management or reporting system.
It is important to note that a number of faults will only reveal themselves in some or even in one machine state and that some faults will only reveal themselves during transient states, such as run-up/coast-down. In conclusion, Classification is essential to improve identification sensitivity, to create realistic prediction and to avoid false warnings.
Predictions are available in the defined machine states with the benefit of giving critical dates for alternative operating conditions. For faults where a combination of parameters is required for diagnosis, the AutoDiagnosis™ function provides fault symptom prediction.
Typical start-up behaviour of a gas turbine Engineers' tools Detailed manual analysis is possible using the signatures plots as well as the "manual analysis tools" such as Orbit, 3-D, Multiple Trend Analysis, Scatter and Nyquist analysis, etc. Statistical analysis functions are available for easy data classification and investigation of relationships between parameters.
Level
The AutoDiagnosis™ fault libraries are expandable to new experience and new machine designs.
1
2
3
Steady States
4
5
Typical Transient and Steady States:
Transient States Gradient States Calibration
The time a machine is not producing is defined as . Production states are defined as . TRANSIENT STATES State 1 Calibration State 2 Slow roll State 3 Ventilation, temperature setting State 4 Run-up (cold and/or hot) State 5 Over speed testing State 6 Synchronization Etc.
6
7 1. 2. 3. 4. 5. 6. 7. 8. 9.
8
9
Time
Test phase Slow roll Ventilation Run-up, cold, warm, hot Over speed testing Syncronisation to mains Low load Medium load .............
Prediction of fault development Prediction of fault development is performed on selective areas in signatures, for which the level relates to specific machinery problems. The system automatically forecasts the date that a fault symptom will reach the defined alarm level. As all data is classified to comparable process circumstances, the prediction provides a reliable estimate.
Extensive data filter and sorting facilities allow fast and easy access to historic data and a wide scope of graphic tools provide presentation in any appearance domain for easy assessment and troubleshooting.
Automatic fault diagnosis (AutoDiagnosis™) OPENpredictor™ provides automatic fault diagnosis for faults in critical machine components and potential faults. The foundation of AutoDiagnosis™ is the synergy between classification, normalization and fault selective signatures. The result is early fault detection and accurate diagnosis with diagnostic messages presented to users in clear text. Locally defined recommendations for action, together with prediction of fault development as well as date for required inspection/maintenance ease the tasks for the users.
OPERATIONAL STATES State 7 Low load, inductive operation State 8 Medium load, inductive operation State 9 High load, inductive operation State 7b Low load, capacitive operation
Fault prediction provides the user with clear information required for maintenance planning or operation improvements.
State 8b Medium load, capacity operation State 9b High load, capacity operation Etc.
6
Any measured signature can be plotted in a 3-D presentation as function of time, RPM or any other process parameter to show dependencies. Applying data filters removes data, which is temporarily not relevant.
The 3-D representation may also be viewed from a birds perspective in the so-called Campbell plot. Here, the Frequency-Time dependent amplitude is shown in colour code.
7
SIGNATURE PROCESSING SYSTEM
The Circular Analysis Plot (CA) is used to monitor changes in symmetry for exhaust gas temperatures, thrust bearing pressure and temperature.
RO2000 Signature Processing Unit (SPU)
The advanced library of signatures provided by the OPENpredictor™ system supported by the very latest technology and performance in signal processing as implemented on the OPENpredictor™ signal processing module, paves the way for detection and diagnosis for a wide variety of all mechanical machine faults. Only a limited number of sensors are required, of which most may already be available on the machines to be monitored. Signals from field sensors are routed to the OPENpredictor™ signature processing unit (SPU), which consists of up to ten signal-processing modules, each handling eight sensor inputs. Processing in the SPU is performed by the individual signal processing modules. Parameters and signatures (including alarms) are transferred to the OPENpredictor™ server for AutoDiagnosis™ processing, prediction, reporting and display. All results are stored in the OPENpredictor™ database for retrieval later, as required.
Signature Processing Unit (SPU) The SPU will be configured from the OPENpredictor™ server. The SPU performs data acquisition, signal conditioning, classification and signature calculation functions. It also performs continuous alarm checks on scalar and vector data as well as on the calculated signatures.
The Shaft Centre Line Plot (SCL) shows the location of the shaft centre in the bearing clearance. The acceptance location domain is indicated by the yellow/red alarm limits.
Sensors Existing sensors Existing sensors will be used as much as possible. The signal output of existing vibration sensors can be duplicated to the SPU inputs using galvanic isolators or direct-buffered sensor outputs of already installed vibration-monitoring panels. This reduces engineering and installation activities to a minimum. Modulation spectrogram as basis for SMD-function.
Load and Reactive load signals will be directly measured by the SPU for the major operation Classification. RO2020-1 Signal Processing Module (SPM)
Other process signals, e.g. for gas turbine efficiency monitoring will be duplicated using isolators in the marshaling racks or a serial link from the control system.
Scalar dynamic data On-line scalar data are overall vibration values, position values, axial positions and machine speeds.
Additional sensors The machines will be equipped with additional dynamic sensors as needed, in order to calculate all the signatures needed for complete fault detection and diagnosis.
Vector data The SPU processes the raw vibration signals to calculate vector values. These vector values improve the AutoDiagnosis™ functions and provide the basic input to the balancing program (option).
SMD function showing synchronous modulation strength as function of time.
Signature calculation and analysis The signal processing modules acquire data and perform the signature calculations. Signature analysis
RO1000 Accelerometer (OEM product).
8
5
provides a high sensitivity for early fault identification and is used for AutoDiagnosis™ and prediction.
Information retrieval The OPENpredictor™ system has been designed for unmanned operation. As machines are typically operating reliably over a long period of time, it is essential that a PMIS system automatically provide warnings, when actions need to be defined.
OPENpredictor™ server
The Cepstrum Signature (CEP) provides a good selectivity to identify gear wheel faults in complex gearboxes and any other faults, which result in a signal with modulation character.
The Selective Envelope Detection Signature (SED) is specifically sensitive to identify faults where symptoms are of repetitive impulse character, such as rolling element bearing faults and cavitation. The AutoDiagnosisTM function uses the repetition rate of the fault to conclude the origin of the fault.
The OPENpredictor™ server receives the data measured and processed by the SPU's. The server stores the data in the internal RAM memory, from which it is transferred into the relational database. In this way large bursts of data will not result in data loss. If data variations are less than a user-defined percentage from previously stored values, then data is automatically removed. This assures that the database is not filled with data, which does not provide any valuable information. Alarm information and fault diagnostic results are also stored in the database.
There • Via • Via • Via • Via
Operation via AutoDiagnosis™ messages OPENpredictor™ automatically interprets changes in machine behavior and will issue AutoDiagnosis™ messages when machinery faults have been positively identified.
The OPENpredictor™ server also runs the AutoDiagnosis™ and health prediction programs, data display and reporting programs as well as the remote access program.
A distinction is made between instantaneous and gradually developing faults. An instantaneous fault does not have a prediction e.g. a shaft rub or cavitation, as it occurs during a certain period as a coincidence of operation circumstances.
Man Machine Interface The user The man machine interface has been developed to meet different user requirements. Users are typically operators in the control room, maintenance engineers, process engineers and system managers. As these users have different responsibilities and demands, the system can be configured to meet individual information requirements.
The Transient Signature (TRT) is used to automatically identify changes in the run-up or coast-down behavior of a machine. Changed critical frequencies and decreased damping are automatically diagnosed.
Vector Analysis is used extensively in the AutoDiagnosis models. OPENpredictor™ provides a unique implementation of Order Tracking Analysis (OTA) that provides accurate vector values, both in steady state and transient conditions. A vector consists of both an amplitude and phase value, referencing to the position of the shaft at the tacho probe. The unique algorithm provides the vector values for four harmonic components and the sub-harmonic representing shaft instability.
Wear phenomena in bearings and gears however typically have a long development path, for which a prediction will be presented. The AutoDiagnosis™ message pops up automatically and provides information about the identified fault, the component involved, the fault strength and if configured by the plant a recommendation of the action to be performed. If a prediction is available then the fault strength can be plotted as a trend together with its expected future development to assist in planning of the corrective task.
OPENpredictor™ incorporates three different warning systems, one each for operation, maintenance and process optimization. System operation is similar for each user but the type of information presented can be adapted to the plant and user requirements.
Alarm Bar
4
exists four different ways to operate the system: AutoDiagnosis™ messages the warning systems the mimics the browser
9
Fault diagnostic messages shall be acknowledged by the user and can simply be printed or routed to the e-mail system.
function minimizes machine operation risk, as any important identified change in machine behavior will give a warning. When OPENpredictor™ has identified an "alert" or "alarm", a "warning lamp" starts flashing, and by clicking the flashing "warning lamp" an "alert list" will show detailed information. One more mouse click reveals the information either as a graphical plot or as an AutoDiagnosis™ message. Accessing data is extremely simple using well-known icons and menu-bars.
Fault Diagnostic Message.
From the fault diagnostic message, the user can access a prediction plot (see example below), forecasting how the fault will develop over time. Operation via warning system This is a very efficient way to retrieve information about identified changes in the machine behavior, for which no AutoDiagnosis™ can be provided yet. This
The Prediction curve presents graphically the specific fault Prediction, together with the confidence range.
can only be recovered by replacement of internal gas turbine components.
(residual lifetime of concerned components). The following illustrates some of the signatures used in OPENpredictor™ together with their main fault identification capabilities.
To monitor the deviation from a new gas turbine performance, is valuable to schedule compressor cleaning. The cleaning and filter exchange date can be forecasted and performed on the most economical date. Additionally, the long-term retrofitting can be economically optimized. The performance monitoring system uses existing process measurements such as temperatures, pressures, fuel specification, ambient conditions, etc. The system provides actual and corrected (corrected to reference ambient) estimates on: • Power output • Heat rate, thermal efficiency • Compressor efficiency • Turbine efficiency • Fouling indices for compressor and filter Process problems that influence machine availability Potential dangerous process circumstances can cause machine damage such as: • Cavitation • Surge • Stall • Combustion pulsation
The Constant Percentage Bandwidth Signature (CPB) gives a general health overview of the majority of mechanical fault characteristics, such as, unbalance, misalignment, foundation looseness. It is however, specifically sensitive in identifying faults such as rolling element bearing faults, cavitation, combustor resonance and gas leaks.
When these problems occur, the operation shall be changed in order to avoid strong dynamic forces to blades, impellers and seals. The result is reduced operational risk and longer service life.
OPENpredictor™ machine health assessment methodology Early fault identification OPENpredictor™ provides a unique library of signatures dedicated to detect and identify most mechanical problems, which may be encountered for common rotating and reciprocating machinery. The dedicated signatures check for different fault characteristics and provide selective information regarding fault symptoms, fault development as well as fault locations. Information provided by the different signatures form the basis for automatic fault diagnosis (AutoDiagnosis™) and prediction of fault development
10
The Autospectrum Signature (FFT) is used to identify faults which can only be diagnosed with a high frequency resolution, such as electrical excitation, distinction between electrical and mechanical imbalance, blade passing excitation and system resonance.
3
OPENpredictor™ CONDITION AND PERFORMANCE MONITORING SYSTEMS
Automatic fault diagnosis and health prediction
FAULTS ON ROTATING COMPONENTS Rotor/shaft: Unbalance Bent rotor Eccentricity Oil whirl Steam whirl Rubbing in bearings & seals
To assess the mechanical and functional health of plant assets, the widest range of potential machinery problems shall automatically be identified. During different transient and stationary operational states of the machines, the development of specific faults shall be identified. As the development of faults is related to machine operation, the measured data has to be classified to transient and operational states in order to achieve meaningful data comparison, automatic fault diagnosis (AutoDiagnosis™) and health prediction.
Blades & impellers: Rubbing Cavitation surge Stall
In the following overview, several machinery components, potential problems and their early diagnosis are presented. The OPENpredictor™ system can however be configured to identify a wide range of other machine specific problems when more detailed machinery information is provided.
Casing: Electrical excitation Misalignment Thermal uneven expansion Blocked bearing movement
Operation via mimics Operation via the mimics is also straightforward. Machine pictograms or photos are used to present sensor locations. Operator data is presented below the defined sensors. This data is automatically updated when new values have been identified. The color of the data represents the alarm status. By clicking on a sensor a graphical plot is started showing the trend data.
Operation via browser For the vibration expert or process engineers the browser is an effective way to retrieve any data from the database. Dedicated filtering functions allow selective display of data, which eases trouble-shooting. The experts can pre-configure preferred data displays in order to ease future data retrieval. The window contains four areas below the title bar: • Menus • Toolbars • Current item/measurement selected • Left field: Item/measurement selector • Right field: Item/measurement data
Combustor: Resonance Flame instability Flame distribution
Rolling element bearings: Outer race defects Inner race defects Cage defects Lubrication deficiency
Reporting Report generator OPENpredictor™ incorporates a report generator to create dedicated reports. These reports can be stored under user-defined names for easy retrieval. Apart from these reports any graphical plot can be exported in HTML format for easy transfer via an e-mail system.
Journal bearings: Shaft lift fault Wear
Machine components The overall health of a machine depends on the condition of all components critical for the machine operation and all dynamic forces acting on the components. The components are subdivided into rotating and stationary items such as: Mechanical and functional health assessment ROTATING COMPONENTS Rotor/Shaft Gear wheels Blades Couplings
FAULTS ON STATIONARY COMPONENTS Foundation: Looseness
Thrust bearings: Wear Couplings: Locking Gear wheels: Wear Back lash Tooth defects Pitting
STATIONARY COMPONENTS Foundation Casing Seals Bearings Combustor
Example of Machine Mimic with real time measurement values and links to other mimics and graphical plots.
Shift reports Shift reports will be printed out automatically at the specified time, after they have been defined (once). They will contain the requested operation information for the shift group in the control room, stating where special attention is needed.
It is possible to create different levels of mimics, which are actively linked together via "link buttons". This allows easy navigation between machines located in different plant sections without having detailed knowledge about machine names, tag identifiers and sensor locations.
Management reports Management reports have to be requested manually. These reports typically will contain maintenance information on identified machinery faults, fault predictions and recommendations. For the experts detailed data can be included to justify the maintenance activities.
For specific machinery design, additional potential faults can be specified.
When no design and/or installation problems exist, and then the typical sources of machine health deterioration are component fatigue, wear, erosion and external factors. Machine mechanical health monitoring provides information to optimize machine availability, while machine functional health assessment provides information with regards to machine performance (e.g. component efficiency).
Monitoring of e.g. gas turbine performance degradation Gas turbine performance degrades over time due to recoverable and non-recoverable mechanical changes. The result is a reduction of the power output and an increase of the heat rate (a decrease of thermal efficiency). Typical recoverable losses are fouling of blades (mostly compressor blades) and air filters. These losses can be recovered by operational cleaning procedures or by external maintenance.
Monitoring problems that influence machine availability The following list is an overview of the most common faults to be monitored in order to assess the mechanical machine health:
Typical non-recoverable losses are blade erosion or deviations in tip clearance and seals. These losses
2
The OPENpredictorTM browser provides hierarchical machine/ components/measurement overview and configuration details. From the browser, graphical representations of measurements can conveniently be invoked.
11
OPEN predictorTM
INTRODUCTION TO THE OPENpredictor CONDITION AND PERFORMANCE MONITORING
SYSTEMS
Copyright2005. All rights, title and interest in and the Software, Hardware and Services detailed in this document and all copyrights, patents, trademarks, service marks or other intellectual property or proprietary rights relating thereto belong exclusively to ROVSING Dynamics A/S
Marielundvej 41 · DK-2730 Herlev · Denmark Tel: +45 46 90 72 00 · Fax: +45 44 84 60 40 · VAT No. 20 05 24 73 e-mail: [email protected] · www.rovsing-dynamics.com
CBS_OP-Intro_Jan05.04_03
TM