CN102567569B - The system and method modeled for the mixed risk of turbine - Google Patents

Abstract

The present invention relates to systems and methods for hybrid risk modeling of turbines. Systems and methods for enhancing turbine operation are disclosed herein. Such systems and methods include hybrid risk models (82). The mixed risk model (82) includes physically based submodels (78, 102, 206) and statistical submodels (80, 208). The physics-based submodels (78, 102, 206) are configured to model physical components of the turbine (10). The statistical sub-model (80, 208) is configured to model historical information of the turbine (10). The hybrid risk model (82) is configured to calculate turbine parameters (84, 86, 88, 202, 204, 208).

Description

Systems and methods for hybrid risk modeling of turbines

technical field

The subject matter disclosed herein relates to systems and methods related to risk modeling.

Background technique

Various systems, such as turbine systems, can include complex mechanical interrelationships between various components and subcomponents. For example, a turbine may include one or more rotor stages (eg, wheels and blades) capable of axial rotation. The individual stages of blades or vanes are capable of converting fluid flow into mechanical motion. The buckets are attached to the rotor wheel by various fasteners such as lockwires. Unfortunately, fasteners can exhibit wear (eg, stress cracking) and require repair or replacement. Similarly, other components of the turbine system may exhibit wear and require repair or replacement. Currently, manual inspection and testing procedures are used to determine whether a component should be repaired or replaced. Such inspections and tests require turbine system shutdown, which is typically time consuming and expensive.

Contents of the invention

Certain embodiments commensurate in scope with the originally claimed invention are summarized below. These embodiments are not intended to limit the scope of the claimed invention, but rather these embodiments are intended only to provide a brief summary of possible forms of the invention. In fact, the present invention may encompass a wide variety of forms that may be similar to or different from the embodiments set forth below.

In a first embodiment, a system for analyzing a turbine includes a hybrid risk model. Mixed risk models include physically based and statistical submodels. The physics-based submodel is configured to model the physical components of the turbine. The statistical sub-model is configured to model historical information for the turbine. A mixed risk model calculates turbine parameters.

In a second embodiment, a non-transitory machine-readable computer medium includes a hybrid risk model. Mixed risk models include physically based and statistical submodels. The physics-based submodels are configured to model the physical components of the turbine system. The statistical submodel is configured to model historical turbine system information. A hybrid risk model calculates turbine system parameters.

In a third embodiment, a method of generating a hybrid risk model includes analyzing physical components of a turbine for a physics-based analysis. The method also includes analyzing statistical information of the turbine to obtain a statistical analysis. Furthermore, the approach includes integrating physics-based analysis and statistical analysis. A hybrid risk model is derived based on the integration of physics-based analysis and statistical analysis. The mixed risk model is configured to calculate turbine parameters.

Description of drawings

These and other features, aspects and advantages of the present invention will become better understood upon reading the following detailed description with reference to the accompanying drawings, in which like symbols refer to like parts throughout the several views, in which:

FIG. 1 depicts a cross-sectional view of one embodiment of a turbine system, illustrating exemplary components;

Figure 2 depicts a detailed view of one embodiment of the components of the turbine system shown in Figure 1;

Figure 3 depicts a flow diagram of one embodiment of modeling and asset management logic;

Figure 4 depicts a flow diagram of one embodiment of hybrid risk modeling logic;

Figure 5 depicts a flow diagram of one embodiment of identification logic;

Figure 6 depicts a flow diagram of one embodiment of maintenance factor calculation logic;

FIG. 7 depicts a flow diagram of one embodiment of multiple hybrid risk models; and

FIG. 8 depicts a flow diagram of one embodiment of a process suitable for predicting rotor wheel failure.

Parts list:

10 gas turbine engine

12 fuel nozzles

14 compressors

16 burners

18 air inlet

level 20

Level 22

Level 24

26 guide vane

28 blades

30 impeller

32 axes

34 Turbo

36 diffuser

Level 40

Level 42

Level 44

46 blades

48 impeller

50 impeller

52 impeller

54 axis

60 exhaust

62 Sewing piece

64 external side

66 inner side

68 air cooling slots

70 logic

72 squares

74 running data

76 Monitoring and diagnostic data

78 squares

80 squares

82 mixed risk models

84 squares

86 squares

88 squares

90 squares

92 logic

94 data sources

96 unit data

98 field monitoring data

100 blade configuration data

102 Physical Model Data

104 logic

106 boxes

108 boxes

110 boxes

112 boxes

114 box

116 boxes

118 box

120 squares

122 boxes

124 High Risk Units

126 squares

128 logic

130 databases

132 boxes

134 boxes

136 boxes

138 squares

140 squares

142 boxes

144 boxes

146 squares

148 squares

150 boxes

152 boxes

154 squares

156 squares

158 variables

160 logic

162 logic

164 logic

166 metal properties

168 metal temperature value

170 boxes

172 boxes

174 squares

176 squares

178 squares

180 squares

182 squares

184 squares

186 stats

188 squares

190 boxes

192 boxes

194 boxes

196 squares

198 squares

200 squares

202 box

204 Equivalent Burn Hours Model

206 submodels

208 submodels

210 submodels

212 submodels

214 submodels

216 hold time

218 submodels

220 submodels

222 submodels

224 submodels

225 submodels

226 model

228 submodels

229 Metal temperature transfer function

230 stress

231 distribution

232 submodels

234 submodels

236 submodels

238 submodels

240 submodels

242 submodels

243 ISO-based submodels

244ISO metal temperature

250 logic

252 logic

254 logic

256 model

258 model

260 square frame

262 model

264 model

266 squares

268 squares

270 squares

272 boxes

274 squares

detailed description

One or more specific embodiments of the invention will be described below. In an effort to provide a concise description of these embodiments, not all features of an actual implementation may be described in the specification. It should be understood that in developing any such actual implementation in any engineering or design project, a number of implementation-specific decisions must be made to achieve the developer's specific objectives, such as compliance with system-related and business-related constraints , these constraints may vary from implementation to implementation. In addition, it should be understood that such a development effort might be complex and time consuming, but nevertheless would be a routine undertaking of design, production, and manufacture for those of ordinary skill having the benefit of this disclosure.

When introducing elements of various embodiments of the invention, the articles "a," "an," "the," and "said" are intended to mean that there are one or more of the elements. The terms "comprising," "comprising," and "having" are intended to be inclusive and mean that there may be additional elements other than the listed elements.

The disclosed embodiments include systems and methods for predicting equipment failure, optimizing operational lifecycles, and/or improving maintenance procedures for mechanical systems. More specifically, the disclosed embodiments include generating statistical analyzes that enable integration of physics-based analyzes or models with empirical data observed during real-world use of mechanical machines, such as the turbine system described in more detail below with respect to FIG. 1 Or a mixed risk model of the model. The hybrid risk model also enables unit-level prediction of failure, lifecycle optimization, and/or improved management of individual units such as individual turbine systems. That is, a fleet of turbine systems such as a fleet of MS-7000F turbine systems, a fleet of MS-7000FA turbine systems available from the General Electric Co. of Schenectady, New York and/or a group of MS-9000F turbine systems, can be operatively managed on individual turbine stages, allowing individual management of substantially all turbine devices in the group. Furthermore, embodiments described herein allow sharing of data, models, calculations, and/or processing across groups of turbines, thereby enabling multi-level operational management of groups of turbines (eg, unit level and group level).

Statistical analysis may be used, for example, in an attempt to predict the failure risk of turbine components based on historical data. However, this statistical analysis may not be accurate, especially when applied to unit-specific forecasts. Physics-based analysis of components may also be used in an attempt to predict equipment failure. Such physics-based analysis can generate models that include virtual representations of components. This virtual representation can then be used, for example, to simulate "wear and damage" of components. However, such physics-based analysis alone may not achieve the desired level of predictive accuracy. The disclosed embodiments allow the derivation of hybrid risk models that integrate certain statistical and physics-based analyses. This mixed risk model can lead to improved prediction accuracy. In fact, the disclosed embodiments allow for an even higher improved level of predictive accuracy over the lifetime of an individual turbine plant or other turbine.

In certain embodiments, the behavior of a particular turbine system can be observed during the operating life of the system, and such observations can be used to predict the occurrence of undesired maintenance events, such as cracks in the lockwire, which may require unplanned maintenance and/or incur additional costs. In fact, the disclosed embodiments improve the operational life of mechanical systems by analyzing data from such systems, determining the likelihood of unplanned maintenance events, and recommending replacement of certain components in order to minimize or substantially eliminate system Unplanned interruption of operations. Thus, more improved maintenance planning and asset management of the systems in the turbine group may be achieved. In fact, the operating life of the analyzed turbine can be improved while reducing or substantially eliminating the occurrence of unscheduled maintenance events.

It may be beneficial to first discuss some embodiments of mechanical systems that may be used with the disclosed embodiments. With the foregoing in mind, reference is now made to FIG. 1 , which illustrates a cross-sectional side view of one embodiment of a turbine system or gas turbine engine 10 . Mechanical systems, such as turbine system 10 , are subjected to mechanical and thermal stress during operating conditions, which may require periodic maintenance or replacement. During operation of turbine system 10 , fuel such as natural gas or syngas may be directed to turbine system 10 through one or more fuel nozzles 12 and into combustor 16 . Air may enter turbine system 10 through intake section 18 and may be compressed by compressor 14 . Compressor 14 may include a series of stages 20 , 22 and 24 that compress air. Each stage may include one or more sets of stationary guide vanes 26 and blades 28 that rotate to provide compressed air in steps of increasing pressure. Blades 28 may be attached to a rotating impeller 30 connected to a shaft 32 . Compressed discharge air from compressor 14 may exit compressor 14 through diffuser section 36 and may be directed into combustor 16 for mixing with fuel. For example, fuel nozzle 12 may inject a fuel-air mixture into combustor 16 at a suitable ratio for optimal combustion, emissions, fuel consumption, and power output. In certain embodiments, turbine system 10 may include a plurality of combustors 16 arranged in an annular arrangement. Each combustor 16 may direct hot combustion gases into a turbine 34 .

As depicted, turbine 34 includes three separate stages 40 , 42 and 44 . Each stage 40 , 42 and 44 includes a set of blades or buckets 46 coupled to a respective rotor wheel 48 , 50 and 52 which is attached to a shaft 54 . As the hot combustion gases cause turbine blades 46 to rotate, shaft 54 rotates to drive compressor 14 and any other suitable load, such as an electrical generator. Ultimately, turbine system 10 diffuses and exhausts the combustion gases through exhaust section 60 .

Turbine components, such as blades or vanes 46 , may be attached to rotor wheels 48 , 50 , and 52 by fasteners, such as lockwires shown in FIG. 2 . The blades 46 and lockwires are subjected to high temperatures and pressures during engine operation. Periodic inspections may be performed to test and verify that the lockwire and blade 46 are within specified operating parameters. For example, eddy current testing may be used to analyze lockwires, air-cooled slots, outer tang fillets, and inner tang fillets for each blade 46 . However, these tests are typically performed offline with the turbine system 10, which can be very expensive and inefficient.

FIG. 2 shows a detailed view of one embodiment of a rotor wheel (eg, rotor wheel 48 , 50 , or 52 ). Each rotor wheel 48 , 50 or 52 includes fastening means, such as lockwires 62 , adapted to couple the blades 46 to the respective rotor wheel 48 , 50 or 52 . The lockwire 62 includes an outer side 64 facing outward generally away from the center of the rotor wheel 48 , 50 or 52 , and an inner side 66 facing inward generally towards the center of the rotor wheel 48 , 50 or 52 . The rotor wheel 48, 50 or 52 also includes air cooling slots 68 for reducing the temperature of the wheel 48, 50 or 52 during wheel rotation. The lockwires 62 and the air cooling slots 68 may experience unscheduled maintenance events. For example, crack formation may occur at either the outer side 64 or the inner side 66 of the lockwire 62 . Similarly, the air cooling slot 68 may undergo crack formation around its perimeter.

As discussed in further detail below, the disclosed embodiments include generative models, such as hybrid risk models, capable of capturing the physical properties of the analyzed components (e.g., impellers 48, 50, 52) and integrating physics-based models with statistical analysis . Such a unit-level hybrid risk model may be used, for example, to predict the risk of an unplanned event for a specific turbine system 10 in a cluster. The component level hybrid risk model can also be used to predict the risk of unplanned events in a cluster with respect to components and component locations such as the outside side 64 of the lockwire 62 , the inside side 66 of the lockwire 62 , and the air cooling slot 68 . Accordingly, the probability of an unscheduled maintenance event for an individual turbine system or unit 10 based on the actual number of hours fired may be calculated. Furthermore, a hybrid risk model may be used to optimize the operation of individual or all turbine units 10 in a group. For example, more efficient maintenance and downtime planning can be achieved through use of the predictive embodiments described herein. It will be appreciated that the techniques described herein can be used with virtually any mechanical system that experiences "wear and tear". In fact, asset management logic suitable for managing a wide variety of mechanical assets, such as the asset management logic in FIG. 3 below, can be used for many mechanical systems, including turbine system 10 .

FIG. 3 is a flowchart of one embodiment of logic 70 that may be used to model and manage a turbine asset, such as turbine system 10 . It will be understood that the logic 70 and disclosed embodiments may be used with any turbomachine, such as turbines, compressors, and pumps. Turbines may include gas turbines, steam turbines, wind turbines, water turbines, and the like. Additionally, logic 70 may include non-transitory machine-readable code or computer instructions that may be used by a computing device to convert data, such as sensor data, into a hybrid risk model and asset management process. Additionally, the logic 70 and any of the models and sub-models described herein may be stored in a controller and used to control logic and maintenance activities related to, for example, turbines and assets of the turbines. Accordingly, a wide variety of data from each individual turbine system 10 may be collected (block 72 ). Data may include operational data 74 and monitoring and diagnostic (M&D) data 76 . Operational data 74 may include maintenance history for individual units 10 in the group, including maintenance log data, such as hardware configuration history, and dates and types of repairs. Operating data 74 may also include dates and types of turbine starts (eg, warm starts, intermediate starts, and cold starts) and any unscheduled maintenance events (eg, cracked lockwires, cracked air cooling slots). M&D data 76 may include, for example, data transmitted by sensors at numerous locations and systems on turbine 10 , such as on fuel nozzles 12 , compressor 14 , combustor 16 , turbine 34 , and/or exhaust section 60 . Additionally, sensed data may include temperature, pressure, flow rate, rotational speed, vibration, and/or power production (eg, watts, amps, volts).

A physics-based maintenance factor (MF) calculation may be derived for each unit 10 in the group (block 78). In one embodiment, the MF calculation is based on a Life Parameter (LP) function or curve. The LP function is used to define operating life at certain temperatures for a component and/or location in a component, such as lockwire 62 and/or air cooling slot 68 . LP functions can be obtained via physics-based modeling techniques—such as low cycle fatigue (LCF) life prediction modeling, computational fluid dynamics (CFD), finite element analysis (FEA), solid modeling (e.g., parametric and nonparametric Modeling) and/or 3D to 2D FEA mapping to model mechanical components (e.g. blades, lockwires, air cooling slots). In fact, a wide variety of modeling techniques may be used, including thermo-fluid dynamics techniques that result in numerical and physical modeling of the turbine system 10 and turbine components. In one embodiment, the LP function may be derived as a transfer function at various metal temperatures based on the metal's temperature, stress, and burn hours per start (ie, N _ratio ), as described in more detail below. The LP function can then be normalized to produce a Normalized Life Parameter (NLP) function or curve. MF can then be obtained roughly as the inverse of NLP, that is, MF=SSF*1/NLP, where SSF is the stress scaling factor for different member configurations (e.g., curved slots versus square slots), as described in more detail below described.

A data mining activity may be employed (block 80 ), which may use operational data 74 and M&D data 76 as input. Data mining input can be preprocessed and then analyzed to extract patterns from the data. Data mining techniques may include clustering techniques, classification techniques, regression modeling techniques, rule learning (eg, association) techniques, and/or statistical techniques suitable for identifying patterns or relationships in input data. For example, clustering techniques can discover groups or structures in data that are in some sense "similar". Classification techniques may classify data points as certain groups of components, such as turbines 10, have a higher probability of encountering unplanned maintenance events. Regression techniques can be used to find functions that model data within a certain margin of error. Rule learning techniques can be used to find relationships between variables. For example, learning using rules may result in associating certain cold start procedures with increased blade wear. Physics-based MF calculation (block 78 ) and data mining (block 80 ) may enable generation of a multi-location multi-level hybrid risk model 82 .

The multi-location, multi-level hybrid risk model 82 can operate at different levels of the turbine system 10, for example, the model can enable analysis of the turbine system 10 as a whole, for turbine system components such as rotors or compressors, for turbine system components such as rotors or The individual rotor components of the blades, as well as the ability to predict for individual segments of the rotor wheel such as lockwire 62 and air cooling slots 68 . The hybrid risk model can also operate at the location of a system, such as turbine system 10 . Example locations for predicting results may include an intake section, a compressor section, a rotor section, and an exhaust section. Virtually any location or section of turbine system 10 may be used. Additionally, the multi-location, multi-level hybrid risk model 82 enables unplanned event prediction (block 84 ), rotor life optimization (block 86 ), and/or rotor retirement (block 88 ).

Unplanned event prediction (block 84 ) may be used to predict unplanned events, such as lockwire events, air cooling slot events, metal stress related events, temperature stress related events, and/or operational usage related events. That is, the probability of an unplanned maintenance event, such as a split seam, can be predicted for an individual unit 10, and corrective action can be taken before the event actually occurs. For example, burn hours may be used to predict a high likelihood of an unscheduled maintenance event for a particular rotor wheel. Turbine system 10 may then undergo preventive maintenance to inspect and/or replace the rotor wheel. In fact, such predictive capabilities allow for more optimized service life and improved performance of turbines such as turbine system 10 . Thus, the predictive capabilities of the techniques disclosed herein allow for rotor life optimization (block 86).

Rotor life may be optimized, for example, by generating and following maintenance procedures based on the actual usage and life history of the particular turbine system 10 and one or more hybrid risk models 82 (block 86 ). The maintenance program may include previous maintenance history of turbine system 10, component installation history (e.g., type of component installed), hours of operation (including hot start, warm start, and cold start hours), type of fuel burned (e.g., liquid fuels, syngas), resulting loads, operating data 74 and/or M&D data 76 are taken into account. A procedure for predicting rotor retirement (block 88 ), as described in more detail below, may also be used to maximize the use (eg, hours of use) of the rotor prior to retirement and replacement of the rotor.

Asset management (block 90 ) for the turbine system 10 may thus include unplanned event prediction (block 84 ), rotor life optimization (block 86 ), and rotor retirement procedures (block 88 ). The turbine system 10 may further be managed by creating, for example, a computerized system suitable for tracking turbine components and related assets, including the occurrence of planned and unplanned maintenance events, component installation history, operating hours, loads, and other Operational data 74 and M&D data 76 . Such a computerized system may also include a non-transitory computer medium storing hybrid risk model 82 and instructions to update hybrid risk model 82 with new data 74 and 76 . Accordingly, the computerized system may be used at a customer site to manage individual turbine systems 10 or a group of turbine systems 10 . In fact, such a computerized asset management system can increase the operational life of a fleet of turbine systems 10 by continuously monitoring the systems 10, updating the hybrid risk model 82, and enabling better use of managed assets.

FIG. 4 is a flowchart of one embodiment of logic 92 suitable for deriving the hybrid risk model 82 . In the example shown, data input is provided using one or more data sources 94 , such as unit 10 data 96 , OSM (on-site monitoring) data 98 , bucket or blade configuration data 100 , and physical model data 102 . Data sources 94 may include sensors placed on turbine system 10, maintenance logs (e.g., unplanned events, planned events), engineering drawings (e.g., CAD drawings), engineering models (e.g., CFD models, FEA models, solid models, thermal model), and the current turbine system configuration. Data 96 , 98 , 100 and 102 may then be used in physical and statistical analysis logic 104 . Logic 104 may first perform cell data scrubbing (block 106). Cell data cleaning (block 106 ) may preprocess data records, eg, by removing incorrect records and/or duplicate records. Unit data cleaning (block 106) may also convert certain records to include the same units (e.g., metric units, imperial units), normalize time scales (e.g., convert from seconds to fractions), and more generally, Prepare data for further processing. The "cleaned" data may then be used to derive a physics-based lifetime profile (block 108), or LP, as described in more detail below with respect to FIGS. 4-6. After deriving the physics-based lifetime curves, M&D data pre-processing (block 110) may occur, suitable for filtering and cleaning the M&D data. Preprocessing of M&D data is very similar to cell data cleaning (block 106). That is, M&D data preprocessing (block 110) may include removing invalid records, normalizing the data, and preparing the data for further processing.

M&D data preprocessing (block 110) may be followed by task analysis (block 112). Task analysis (block 112 ) may include mathematical and/or statistical analysis of the M&D data 76 and may integrate the MF equations described above with respect to FIG. 3 . Task analysis (block 112) can be used to calculate a set of values for each individual unit 10 in the group, such as midpoint, median, mean, percentage, cumulative distribution function, and/or probability density for a plurality of M&D variables function. A non-exhaustive list of M&D variables may include generator watts (DWATT), turbine horsepower (TNH), fuel reference (FSR), compressor inlet guide vane (CSGV) position, ambient inlet temperature (TAMB), compressor inlet temperature (CTIM), Compressor Discharge Temperature (CTD), Compressor Discharge Pressure (CPD), Compressor Pressure Ratio (CPR), Fuel Stroke Reference Position (FSR), High Pressure Turbine Shaft Speed in % (TNH), Exhaust temperature (TTXM), combustion reference temperature (TTRF1), internal turbine wheelspace temperature before first stage (TTWS1F1) and/or external turbine wheelspace temperature after first stage (TTWS1AO), number of cold starts, number of hot starts, and The number of warm starts. In fact, a wide variety of turbine system 10 values and performance parameters may be used.

The volume of M&D data can be very large, and in some cases data is collected at about five minute intervals over the course of two or more years. Task analysis (block 112) helps identify variables that are particularly suitable for use in the analysis process. Such variables are considered "key X" variables, and the identification logic for such variables is described in more detail below with respect to FIG. 5 . The task analysis (block 112) also extracts or reduces the large M&D data set to selected statistics and Mathematical values (eg, midpoint, median, mean, percentage, cumulative distribution function, and/or probability density function). For example, the mission analysis (block 112) may calculate approximately three-year, two-year, one-year, six-month , the three-month median, midpoint, and/or average, which can be used to calculate equivalent burn hours (box 114).

Equivalent_FH derivation (block 114 ) integrates physics-based model analysis with statistical analysis through the equation Equivalent_FH=MF*FH, where FH corresponds to the actual combustion hours for a given turbine system or unit 10 . In fact, the equivalent burn hours enable individual units 10 to be tracked and managed, and integrate physics-based and statistical MF analysis with the empirical burn hours of each individual unit 10 in the turbine group. Additional statistical techniques, such as correlation analysis (block 116), may be used to process the data as described in more detail below.

Correlation analysis (block 116) may be used, for example, to find relationships between variables suitable for predictive use. In some instances, Pearson correlation analysis can be used to describe the relationship between all M&D factors or variables, and a Pearson coefficient representing the dependence between two variables can be derived and used. Additionally, equivalent burn hours can be correlated to all M&D factors. Furthermore, physics-based correlations may be used, where variables are related to each other based on their corresponding measurement locations and physical properties (eg, component geometry, metal type). Other statistical correlation techniques can be used, such as t-statistics, inter-group correlations, and/or intra-group correlations. Correlation analysis (box 116) and multivariate analysis (box 118) help identify variables that are particularly suitable for use in the forecasting process. Such variables are considered "key X" variables, and the identification logic for such variables is described in more detail below with respect to FIG. 5 .

Multivariate analysis (block 118) may include analysis of variance techniques (ANOVA) and/or logistic analysis. ANOVA can be used, for example, to analyze the variance of a particular variable (eg, M&D data) and divide that variance into variance components based on likely sources of variation. For example, a warm start may result in a larger fraction of the change in equivalent burn hours. Logistic analysis (ie, logarithmic modeling) enables the derivation of the probability of an event occurring by fitting the data to a logistic curve (eg, an S-shaped curve). Other multivariate analysis techniques can be used, such as MANOVA and multiple discriminant analysis, as described below. Suitable variables identified by the "key X" analysis can then be used in a risk modeling analysis, such as Weibull risk modeling (block 120). In some embodiments, Weibull hazard modeling (block 120) can be used to derive a set of proportional hazards models. A proportional hazards model allows the time elapsed until an unplanned maintenance event (e.g., cracked air cooling slot, cracked seam, impeller replacement, cracked blade) to occur with one or more common variables (e.g., M&D factor, etc. effective burning hours) related. For example, increasing warm actuation by a certain percentage increases the probability of an unplanned event at the medial first level. Weibull risk modeling (block 120) may also incorporate interval review measures suitable for analyzing event occurrences between observations, such as between turbine inspections. The interval inspection measure thus enables the derivation of a survival function between two inspection events that can be used to predict the likelihood of an unplanned event occurring.

Therefore, the risk analysis and recommendations (block 122) can use Weibull risk modeling (block 120) and the aforementioned statistical techniques (e.g., equivalent burn hours calculation 114, correlation analysis 116, multivariate analysis 118) to arrive at a mix The risk model 82 groups and determines any high risk units 124 that can be run in the group. In fact, the hybrid risk model 82 and list of high risk units 124 can be deployed to customers (block 126 ) for use in managing turbine operations and assets. The customer can then use the hybrid risk model 82 to improve usage of the turbine system 10 by enabling more efficient and purposeful maintenance planning for the individual units 10 in the group. Such a capability may result in increased longevity and reduced maintenance costs of the units 10 in the group.

FIG. 5 shows one embodiment of "key X" identification logic 128 that enables classification of multiple variables, such as M&D variables, into variables that are particularly suitable for use in the forecasting process. As noted above, the number of M&D variables can be very large, and the amount of data collected for each M&D variable can be collected at intervals (eg, about five minutes) over a span of several years. Thus, the "key X" identification logic 128 enables a reduction in the amount of variables used in the forecasting process. Logic 128 may first use M&D database 130 through data extraction (block 132) to extract data corresponding to M&D variables (including, for example, DWATT, TNH, FSR, CSGV, TAMB, TIM, CTD, CPR, TNH, TTXM, TTRF1, TTWS1F1 and TTWS1AO) data. The logic 128 may then use the extracted data through data filtering (block 134 ) to validate the data and filter the data. Data validation may include removing incorrect data, such as data with negative values when all values should be positive (eg, time values). Similarly, data filtering may remove or filter certain data that may not be useful, such as data points where TNH<95 and DWATT<15. Logic 128 may then use the filtered data through unit statistical analysis (block 136 ) to derive a set of statistics for each unit 10 in the group. Such values may include maximum, minimum, median, midpoint, cumulative distribution function, and/or probability density function. In some embodiments, unit statistical analysis 136 may derive statistics based on data collected every 30 seconds, 1 minute, 5 minutes, 10 minutes, or 30 minutes. Data imputation (block 138) may then impute or assign any missing values, for example by using the median value found in the cell statistical analysis (block 136). For example, any missing CTD, TTWS1F1, or TTWS1A0 values may be assigned (block 138) a median value for each corresponding variable found during unit statistical analysis (136).

Data processing (block 140 ) may then process and derive correlation values based on the M&D database 130 . For example, TTWS1_temp may be derived based on a maximum temperature comparison between two TTWS1 values, such as the two most recent values (ie, the values found at time n and time n+1). Metal temperature calculations (block 142 ) may then use physics-based functions to calculate metal temperatures at various locations in the turbine system 10 . For example, the temperature of a metal such as Inconel (eg, Inconel IN706) may be found for an air slot located in a first stage of a turbine rotor or for a lockwire located in a second stage of the same turbine rotor. In fact, the metal temperature calculation (block 142 ) may be used to calculate the metal temperature at a number of locations in the turbine system 10 . The temperature difference ΔT can be found based on the equation ΔT = T _ACT - T _ISO , where T _ACT is the actual temperature at the turbine location (eg, air cooling slot, lockwire) and T _ISO is the ISO-day temperature . More specifically, the ISO diurnal temperature corresponds to the International Organization for Standardization (ISO) reference temperature typically used for comparison purposes. Such reference temperatures can be found in ISO documents such as ISO document 2314 "Gas Turbine-Acceptance Test".

The temperature difference ΔT may then be processed through a metal temperature filtering process (block 144 ) in order to filter the temperature ranges at different locations. That is, certain temperature measurements outside a given range may not be used, resulting in a temperature range that is useful in deriving other calculations. For example, for values less than -91°F, ΔT may be set to -91°F, and for values greater than 209°F, ΔT may be set to 209°F. Therefore, the metal temperature filtering process (block 144 ) may help reduce outliers.

A normalized life parameter (NLP) calculation (block 146) may then be used to derive a normalized life parameter (LP). As mentioned above, LP is calculated at different metal temperatures for a given material and location. More specifically, LP calculations based on the time remaining before an unplanned event (e.g., cracked air cooling slots, cracked lockwires, impeller replacement, cracked blades) or risk can be used as metal temperature T _metal , at the location of interest Calculated as a function of the stress σ of , the burning hours per start (ie, the N _ratio ). LP can be derived for different locations (eg, air cooling slots, seams) and configurations, for actual temperatures, ISO diurnal temperatures, and modeled (eg, "virtual" temperatures). Configurations may include turbine frame type (e.g. 7F, 7FA, 7FA+, 7FA+e), blade or blade type (e.g. stage 1B, stage 2B), blade design used (original design, new design) and/or wheel Whether the blade is a back-cut vane. By using a set of physics-based modeling techniques such as low-cycle fatigue (LCF) life prediction modeling, computational fluid dynamics (CFD), finite element analysis (FEA), solid modeling (e.g., parametric and non-parametric Modeling) and/or FEA mapping from 3D to 2D, a suitable function LP=function(T _metal , σ, N _ratio ) can be derived. The resulting LP parameters at various temperatures can then be normalized (ie converted to NLP) by using, for example, the equation NLP=LP/ _LPISO .

NLP curves can be plotted by placing NLP parameters in the y-axis and ΔT values in the x-axis of the NLP curve. In one embodiment, the NLP curve may be derived by fitting the scattered points using a nonlinear fit or function for negative ΔT values and an exponential fit or function for positive ΔT values. The resulting NLP curve maps the NLP parameters for any given ΔT. The MF calculation (block 148) can then convert the NLP parameters to MF values by using the equation MF=SSF*1/NLP, where SSF is the stress scaling factor σ. The stress scaling factor σ may vary based on the configuration in use (eg, turbine frame, bucket type, bucket design, and bucket cutout). More details of the MF calculation - including variations of the MF calculation for units 10 with mixed hardware configurations - are described below with respect to FIG. 6 .

Equivalent Fire Hours Calculation (Block 150 ) The Equivalent Fire Hours (Equivalent_FH) may then be calculated based on the equation Equivalent_FH=MF*FH, which corresponds to the actual Fire Hours for a given unit 10 . Correlation analysis may then be performed (block 152), as described above with respect to FIG. 4, including using ANOVA techniques and/or logistic analysis (block 154). Correlation analysis 152 may use statistical and/or physically based correlations to map relationships between different variables in M&D data 76 . In one example, the logic 128 may classify the data using data mining classification techniques, such as Quadratic Discriminant Analysis (QDA) classification (block 156). For example, QDA classification (block 156 ) may classify the data based on correct failure prediction, incorrect failure prediction, correct failure pending (eg, system halted), and incorrect failure pending. The QDA classification (block 156) can thus be used for comparative measures (eg, ANOVA) for multivariate risk modeling. The result of using the foregoing techniques is the identification of one or more "key X" variables suitable for use in predicting unplanned events. For example, Equivalent_FH, Startup, and Percent Warm Startup can be used as "key X" variables 158 to better predict internal seam cracking at stage 1W. Similarly, the Equivalent_FH and N _ratios can be used as "key X" variables 158 to better predict internal seam cracking at state 2W. It will be appreciated that other statistical techniques may be used to implement the "key X" variable 158 - for example using any suitable correlation analysis, including other forms of multivariate analysis (e.g., MANOVA) and/or suitable discriminant analysis techniques (e.g., linear discriminant analysis, Regularized QDA).

FIG. 6 shows a more detailed view of one embodiment of the MF calculation logic 148 shown in FIG. 5 . In the illustrated embodiment, MF calculation logic 148 may be further subdivided into MF transfer function calculation logic 160 , actual metal temperature calculation logic 162 , and hybrid hardware configuration logic 164 . MF transfer function logic 160 may enable the derivation of a set of LP functions suitable for calculating the base MF _b , while actual metal temperature calculation logic 162 may be used to calculate _ΔTactual . The base MF _b can then be used to obtain the MF of each individual unit 10 . Accordingly, the specific configuration of each turbine system 10 may be taken into account, including configurations having mixed hardware via the mixed hardware configuration logic 164 . Configure hybrid hardware configurations that may have been retrofitted, for example with newer component designs. In fact, the MF calculation logic 148 described herein enables MF calculations for individual turbine systems 10 having a mix of original and updated hardware configurations.

During physics-based modeling (block 170) of turbine 10 components and/or locations (e.g., internal lockwires), the MF transfer function logic 160 may use the metal properties 166 in addition to the ISO date and metal temperature values 168 , such as metal type and material composition. As mentioned above, physics-based modeling (block 170) can be achieved by using methods such as LCF life prediction modeling, CFD, FEA, solid modeling (eg, parametric and non-parametric modeling), and/or 3D to 2D The technique of dimensional FEA mapping derives LP based on the T _metal , σ, and N _ratios as a function. The LP for a number of " _virtual " temperatures Tvirtual (ie, LP _T ) and the LP for ISO diurnal temperatures (ie, LP _ISO ) may then be calculated (block 172 ). The term "virtual" temperature is used to indicate a range of temperature values, which may include the actual measured temperature. For example, the term may indicate a temperature sequence beginning at -10°F and ending at 1200°F with 1°F increments (i.e., -10°F, -9°F, -8°F, ..., 1200°F) for all temperatures. This calculation allows the derivation of normalized LP _T (NLP _T ) by using the equation NLP _T =LP _T /LP _ISO (block 174). _ΔTvirtual may then be calculated (block 176) based on the equation _ΔTvirtual = _Tvirtual -T _ISO (block 178).

The NLP _T and ΔT _dummy values can then be used as part of the data fitting process (block 180), wherein the NLP _T and ΔT _dummy values are set as scattered points such that the NLP _T values are in the y-axis and the ΔT _dummy values are in the x-axis middle. In one embodiment, the NLP _T can be fitted with respect to (the relationship of) the ΔT _dummy by using a nonlinear fit or function for negative or zero ΔT _dummy values and an exponential fit or function for positive ΔT _dummy values The points are scattered to derive (block 180) the transfer function. That is, x-axis values less than or equal to zero are fitted using a nonlinear fit, while positive x-axis values are fitted using an exponential fit.

The NLP is calculated for all ΔT _actual values (block 182) and the resulting transfer function can then be used. As depicted, the ΔT _actual value may be calculated using actual metal temperature calculation logic 162 . As described above with respect to FIG. 4, logic 162 may perform task analysis (block 184). Task analysis can result in a set of values 186 based on statistical performance. The _actual temperature Tactual may be calculated (block 188 ), for example, based on the metal temperature transfer function for various locations and/or component parts and using the property value 186 as input. The derived metal temperature function is thus suitable for calculating the actual temperature of the metal at a particular location (eg, lockwire, air cooling slot) based on the M&D data 76 . The ΔT _actual value may then be calculated (block 190 ) by using the equation ΔT _actual = T _actual - T _ISO .

MF _B may then be calculated (block 192 ) based on the equation MF _B =1/NLP _T . MF _B alone may well be used to anticipate unplanned events (eg, in cases where the underlying hardware is configured using a standard configuration (eg, default installation configuration)). However, some units may be modified eg by replacing components such as rotor blades with components of a more recent design (eg back cut rotor blades). Accordingly, MF _B may be modified by the hybrid hardware configuration logic 164 to take the hybrid hardware configuration into account.

The hybrid hardware configuration logic 164 may extract the hardware and software configuration of each unit 10 (block 194), including a historical listing of the configurations used by each unit 10 and the runtime of each configuration i. The run time ratio RT _i for each configuration i can then be calculated (block 196 ) based on the equation RT _i =T _i /FH, where T _i is the time the configuration is active and FH is the total burn hours of the unit. Stress scaling factors SSF _{i may then be calculated (block 198 ) for each configuration i} . SSF _i takes into account stresses specific to configuration i based on eg metal type, component geometry and/or location. The hybrid configuration SSF can then be calculated (block 200 ) by using the formula SSF=Σ(RT _i *SSF _i ). Thus, mixed hardware configurations may be taken into account by calculating (block 202) MF using the equation MF = MF _B *SSF. Such calculations enable the predictive technique to be applied to substantially any turbine system 10 , regardless of configuration type or configuration installation date.

FIG. 7 depicts an embodiment of a hierarchical mixture risk model 82 . The depicted embodiment includes an Equivalent Burned Hours model 204 (ie, Equivalent_FH=MF*FH), which enables the integration of empirical and physics-based analyzes suitable for calculating the time until an unplanned maintenance event occurs. The hybrid risk model 82 may include a MF calculation submodel 206 and a burn hours submodel 208 . The burn hours sub-model 208 enables the calculation of the observed burn hours in a given unit 10, which may include data cleaning and validation techniques suitable for removing erroneous and invalid data from the observed burn hours. The MF calculation submodel 206 enables calculation of MF (eg, SSF*1/NLP) based on, for example, the NLP submodel 210 . In this embodiment, the NLP submodel 210 may include an actual equivalent FH submodel 212 adapted to calculate actual equivalent burn hours using approximately two years' worth of M&D data (ie, Equivalent — FH _2YR ). It will be appreciated that other embodiments may use smaller or larger data timelines, such as 6 months, 1 year, 1.5 years, 2.5 years, or 4 years. The NLP submodel 210 may also include an ISO-equivalent FH submodel 214 suitable for calculating ISO daily equivalent burn hours (ie, Equivalent_FH _ISO ). Therefore, the NLP sub-model 210 can calculate the NLP value by using the equation NLP=Equivalent_FH _2YR /Equivalent_FH _ISO .

The actual Equivalent_FH submodel 212 can calculate the Equivalent_FH _2YR value by using the equation Equivalent_FH _2YR = N _{i, HT - 2YR} * Hold_Time (hold time), where N _{i, HT - 2YR} is the holding time for a given cycle (HT ) 216 or dwell time, initial life or number of cycles until the onset of an unplanned event such as cracking in the metal. In other words, N _{i, HT-2YR} measures the number of cycles based on hold or dwell time 216 at a temperature during which a location or component with a particular metal type (e.g., Inconel IN706) can Started to crack. N _i,HT-2YR may be calculated by looping through to the initial submodel 218 as a function of _HT 216 , the time-dependent parameter PT , the fatigue parameter P _FAT , and the LFC parameter N _i,20CPM of the continuous cycle.

The cycle to initial submodel 218 uses hold time 216 , cyclic fatigue submodel 222 , low cycle fatigue submodel 224 , and time dependent parameter submodel 224 to derive the calculations performed. Models 220, 222, and 224 are included in an actual sub-model 225 that uses actual data rather than ISO-based data. Hold time 216 is a measure of the amount of time spent in a hold or dwell period. The cycle fatigue sub-model 222 can calculate the fatigue parameter P _FAT based on the equation P _FAT =1/N _i,20CPM , where N _i,20CPM is derived from the low cycle fatigue sub-model 224 . The low cycle fatigue submodel 224 may, for example, yield Ni,20CPM at 20 cycles per minute (CPM) as a function of the uniaxial strain range Δε for a given temperature and metal (eg, Inconel _IN706 ). It will be appreciated that other CPM values may be used, such as 5 CPM, 15 CPM, 25 CPM, 30 CPM, and the like.

The time-dependent parameter submodel 220 enables the calculation of the time-dependent parameter PT _. PT is a parameter suitable for measuring the time until damage occurs and can be obtained based on the metal temperature transfer function 229 described in more detail above with respect to FIGS. 3 and 4 , which in turn uses the M& _D distribution 231 derived from the mission analysis 112 . In one embodiment, the time-dependent parameter PT may also include a mid-life “Neuberized” stress model 226 . That is, to account for the elastic stress concentration factor between the strain factor K _σ and the stress factor K _ε Neuber's rule for the relationship can be used by the stress model 226 . Cyclic fatigue submodel 222 may also include strain range Δε submodel 228 , which may be based on elastic stress 230 . That is, the strain range Δε can be derived by submodel 228 as a function of temperature and elastic stress 230 . The ISO-equivalent FH submodel 214 may include a set of submodels 232 , 234 , 236 , 238 , 240 , 242 that are substantially similar to the submodels 220 , 222 , 224 , 226 , 228 . However, sub-models 232, 234, 236, 238, 240, and 242 use ISO metal temperature 244 instead of actual temperatures. Thus, submodels 234, 234, 236, 238, 240, and 242 are included in ISO-based submodel 243 that uses ISO data rather than just actual data.

More specifically, the time-dependent parameter submodel 234 uses the metal temperature transfer function and the ISO metal temperature 244 to calculate parameters suitable for measuring the time until damage occurs. The cycle fatigue sub-model 236 can obtain the fatigue parameter P _ISO-FAT =1N _i,20ISO-CPM , wherein N _{i, 20ISO-CPM} is obtained from the low cycle fatigue sub-model 238 . The low cycle fatigue submodel 238 may, for example, yield Ni at 20 cycles per minute (CPM) for a given ISO metal temperature 244 and metal (e.g., Inconel IN706) as a function of the uniaxial strain range Δε _{, 20ISO-CPM} . Similarly, the strain 240 and stress 242 submodels may derive strain and stress based on a given ISO metal temperature 244 .

FIG. 8 depicts logic 250 suitable for predicting the total amount N of rotor wheel scrapping by applying the hybrid model described above with respect to FIG. 7 . Logic 250 may be further subdivided into unit-level analysis logic 252 and component-level analysis logic 254 . Unit-level analysis logic 252 may include a unit-level risk model 256 suitable for predicting the occurrence of unplanned events. The unit level risk model 256 may use the hybrid risk model described above with respect to FIG. 7 and may be used to predict the probability of occurrence of an unplanned maintenance event for an individual unit 10 . The unit level risk model 256 may include, for example, the equivalent burn hour mixing model 204 derived for a specific location in the turbine component, such as the inner side of the lockwire. Similarly, a second element level risk model 258 modeling different locations in the turbine component, such as the outer side of the lockwire, may also be used. Accordingly, the second unit-level risk model 258 may also include modeled locations as described with respect to FIG. 7 (eg, Example of a mixed risk model for the outer side of the lockwire).

A prediction of the risk of the unit 10 , such as failure due to lockwire cracking (either internal or external), may then be derived (block 260 ), eg, based on the unit-level risk models 256 and 258 . The prediction of risk (block 260) may include a proportional hazards model (PHM), such as a Weibull PHM, which is adapted to relate certain variables (e.g., Equivalent_FH, N _ratio , warm start percentage) to the combustion rate prior to an unplanned maintenance event. hour dependent. For example, the Weibull PHM may enable the derivation of the probability of occurrence of various unplanned events based on the current burn hours of a given unit 10 .

The component level analysis logic 254 may incorporate a component level risk model 262 adapted to model risks associated with specific components and component locations. For example, a component-level risk model 262 may be derived to model the interior side of the lockwire. In other words, the component-level risk model 262 is similar to the unit-level risk model 256 , but involves modeling the risks of the location of common components rather than the risks associated with the use of components in individual units 10 . Similarly, a component-level risk model 264 may be derived to model risks associated with different locations of a common component, such as the exterior side of a lockwire. Models 262 and 264 may then be used to predict the number of split lockwires (block 266). In one embodiment, the prediction of the number of cracked lockwires (block 266) may include using the probability function Pr(i) derived from models 262 and 264, where Pr(i) is the number of cracked lockwires for individual piece i. probability. Therefore, a set of probability functions {Pr(1), Pr(2), . . . Pr(i), .

In the illustrated embodiment, a Monte Carlo simulation (block 268 ) is used to predict the probability Pr of reaching or exceeding a certain impeller failure threshold (≥retirement threshold). For example, if three or more adjacent lockwires are cracked, the impeller scrapping threshold may be met or exceeded. Any suitable Monte Carlo simulation can be used, including simulating the set of probability functions {Pr(1), Pr(2), . . . Pr(i), . . . , Pr(the total number of slices )} to compute an iterative simulation of the probability distribution. For example, during each iteration, the set of probability functions {Pr(1), Pr(2), ... Pr(i), ..., Pr (total number of slices)} can be used to compute and store a set of probability value. As more iterations of the simulation are performed, the stored values are used to define the probability Pr (≥rejection threshold). Accordingly, the probability of impeller failure may be derived based on summation of all simulation iterations or scenarios (block 270 ).

In one embodiment, a probability of impeller failure rate may be calculated (block 272 ) based on actual inspection results. For example, inspection logs may be analyzed to determine the ratio of actual failures to predicted failures. The probability of impeller scrapping rate (block 272) may then be integrated with the derived impeller scrapping probability (block 270) to calculate the total amount of impeller scrapping (block 274). In fact, by applying the techniques described herein, including the use of a hybrid risk model, maintenance can be significantly improved by enabling predictions of the number of impellers that may need to be retired. For example, obtaining a replacement rotor wheel from the manufacturer may require some lead time or wait time. Accordingly, a parts purchase or parts replenishment system may order a replacement impeller prior to actual retirement. It will be appreciated that the techniques described herein may be used in other applications, such as financial and/or decision support applications. With a dramatically improved set of techniques available for unplanned event prediction, financial decisions that integrate business operations and engineering analysis can now be made. For example, business operations related to inventory management, parts procurement, logic, maintenance planning, maintenance operations, etc. may be improved.

Technical effects of the invention include modeling techniques that enable the integration of physically based modeling and statistical techniques into hybrid models. Hybrid models can lead to improved predictive estimates of events such as unplanned maintenance events.

This written description uses examples to disclose the invention, including the best mode, and also to enable any person skilled in the art to practice the invention, including making and using any devices or systems and performing any incorporated methods. The patentable scope of the invention is defined by the claims, and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims if they have structural elements that do not differ from the literal language of the claims, or if they include equivalent structural elements with insubstantial differences from the literal language of the claims .

Claims (10)

1. A system for analyzing a gas turbine system comprising: a processor programmed to: performing a hybrid risk model comprising a physically based submodel and a statistical submodel, wherein the physically based submodel is configured to calculate The physical components of the gas turbine system are modeled, where the arguments of the life parameter function F include the _metal temperature Tmetal at the location of interest of the gas turbine system, the stress σ, and the combustion hours per start of the gas turbine system, and the The life parameter function F is used by the processor to derive a physically based maintenance factor MF by the formula MF=SSF*1/NLP, where SSF is the stress scaling factor, NLP is the normalized life parameter function F, and the statistic The sub-model is configured to model the historical information of the gas turbine system by calculating actual firing hours for the gas turbine system, and wherein the processor is configured to combine the derivation of MF with the actual firing hours and The equivalent combustion hour parameter Equivalent_FH is calculated by the equation Equivalent_FH=MF*FH, where FH is the actual combustion hour, and the probability of occurrence of the unplanned event is predicted based on the current combustion hour of the gas turbine system, and will be equal to The effective burn hours parameter is converted into a probability of failure of a component of the gas turbine system. 2. The system of claim 1, wherein the processor is configured to determine remaining operating life by using the obsolescence probability. 3. The system of claim 2, wherein the processor is configured to data mine sensor data acquired from a group of gas turbines by performing regression analysis and/or classifying a plurality of data points To obtain a life parameter function F, the regression analysis includes a non-linear fit to a plurality of data points included in the sensor data, the plurality of data points being classified as having the State the expected probability of being a member of the group for the remainder of the time. 4 . The system according to claim 1 , wherein the scrapping probability comprises a scrapping probability of a locking wire, a scrapping probability of an air cooling slot, a scrapping probability of an impeller, a scrapping probability of a blade, or a combination thereof. 5. The system of claim 1, wherein the statistical sub-model includes gas turbine system component installation history, gas turbine system component usage history, gas turbine system group usage history, multiple monitoring and diagnostic sensors data, or a combination of them. 6. The system of claim 5, wherein the statistical sub-model comprises a Weibull risk model configured to measure between the first inspection event of the gas turbine unit and the A survival function is derived between the second inspection events of the gas turbine unit. 7. The system of claim 1, comprising an asset management system, wherein the asset management system collects turbine system data and uses the hybrid risk model and the collected turbine system data to manage the turbine system member. 8. A method of generating and using a hybrid risk model comprising: extract data related to the operation of the gas turbine system; Physics-based analysis is obtained by transforming the operationally relevant data by analyzing the physical components of the gas turbine system to derive the maintenance factor MF by the formula MF=SSF*1/NLP, where SSF is the stress scaling factor and NLP is the normalization The life parameter function F of the life parameter function F, wherein the maintenance factor MF is based on the formula life parameter function F = data-based remaining time before the unplanned event, where the independent variable of the life parameter function F is included at the location of interest of the gas turbine system The _metal temperature Tmetal, the stress σ and the firing hours per start of the gas turbine system; Analyze the historical data of the gas turbine system to obtain the actual combustion hours of the gas turbine system; Integrating the physics-based analysis and the actual fired hours of the gas turbine system into the hybrid risk model, which involves combining the derivation of MF with actual fired hours and obtaining by the equation Equivalent_FH=MF*FH The equivalent combustion hour parameter Equivalent_FH, where FH is the actual combustion hour; and converting the equivalent fired hours parameter into a probability of failure of a component of the gas turbine system by predicting the probability of occurrence of said unplanned event based on current fired hours of the gas turbine system; Wherein converting said operation-related data and converting equivalent combustion hours parameters are performed by computing means. 9. The method of claim 8, comprising deriving the lifetime parameter function F by data mining sensor data extracted for a set of gas turbine classes. 10. The method according to claim 8, comprising identifying the values derived from the set of monitoring and diagnostic variables based on monitoring and diagnostic data obtained from a set of gas turbine classes and by using quadratic discriminant analysis classification. A subset of monitoring and diagnostic variables useful for said probability of failure of said component, wherein converting an equivalent burn hour parameter to said probability of failure of said component comprises combining an equivalent burn hour parameter with said subset of monitoring and diagnostic variables A variable combination then converts the equivalent burn hours parameter with the subset of monitoring and diagnostic variables into the probability of obsolescence.