Patent Application
20180173824
The following relates to a method for performing a model-based failure analysis of a complex industrial system such as a gas turbine system.
A complex industrial system can comprise a plurality of hardware and/or software components. The performance of a complex industrial system depends on operational conditions of the employed components. For reliability assessment, it is important to predict a failure impact of a failure of a component of the system on the functionality of the system in order to assess, whether this can lead to a critical situation if safety or reliability requirements are violated. Further, the prediction of a failure impact can form the basis for measures to minimize or mitigate the failure impact by design correction and/or maintenance of the respective system. Each complex system can have different operating and process requirements and therefore often differs in its specific design. The failure mode and effects analysis, FMEA, can be used to systematically analyze postulated component failures and to identify the resultant effects on system operations. Conventionally, the FMEA analysis is performed and redone for each variant or version of the investigated industrial system and for each revision of a system design. This analysis is often performed by groups of experts being labour- and time-intensive.
An aspect relates to providing automatically fault effect associations which can be used for diagnostic tasks such as root cause analysis.
The following provides according to the first aspect of embodiments of the present invention a method for performing a model-based failure analysis of a complex industrial system consisting of hardware and/or software components each represented by a context independent component model comprising interface terminals and a set of component behaviour modes including a normal mode and failure modes of the respective component stated as constraints on deviations, the method comprising the steps of:
generating a system model of an investigated industrial system by loading component models of the components of said investigated industrial system from a component library and connecting the interface terminals of the loaded component models according to a structure of the investigated industrial system, and
executing a constraint-based predictive algorithm on a reasoning engine to generate qualitative FMEA results for different operation scenarios of the investigated industrial system.
In a possible embodiment of the method according to the first aspect of embodiments of the present invention, the constraint-based predicted algorithm iterates over a Cartesian product of predefined operation scenarios and failure modes of each component to determine, whether the failure propagation entails a local or a system level effect capturing a violation of a functionality of the investigated industrial system.
In a further possible embodiment of the method according to the present invention, the interface terminals of a component model are formed by channels to other components comprising interface variables exchanged with the other components of the investigated industrial system.
In a further possible embodiment of the method according to the present invention, the component model of a component comprises state variables indicating a state of said component.
In a further possible embodiment of the method according to the present invention, the component model of a component comprises a base model capturing a physical behaviour of said component.
In a further possible embodiment of the method according to the present invention, the component model comprises deviation models capturing deviations of actual values of variables from reference values of the variables.
In a further possible embodiment of the method according to the present invention, the component model comprises local effects indicating effects of component faults of said component on a functionality of the investigated industrial system.
In a further possible embodiment of the method according to the present invention, the generated FMEA results are used to predict a failure impact of a failure on the functionality of the investigated industrial system.
In a further possible embodiment of the method according to the present invention, the system model is generated by connecting the interface terminals of loaded component models by a model editor according to a predetermined topology of the investigated industrial system.
In a further possible embodiment of the method according to the present invention, the constraint-based predictive algorithm is executed on said reasoning engine offline during design, maintenance and/or repair of the investigated industrial system and/or online during operation of the investigated industrial system.
In a further possible embodiment of the method according to the present invention, at least one component of said investigated industrial system is controlled in response to the generated FMEA results.
The following provides according to the second aspect of the present invention an apparatus for model-based failure analysis of a complex industrial system consisting of hardware and/or software components each represented by a context independent component model comprising interface terminals and a set of component behaviour modes including a normal mode and failure modes of the respective component stated as constraints on deviations, said apparatus comprising:
a generation unit adapted to generate a system model of an investigated industrial system by loading component models of the components of said investigated industrial system from a component library and connecting the interface terminals of the loaded component models according to a structure of the investigated industrial system, and
a reasoning engine adapted to execute a constraint-based predictive algorithm to generate FMEA results for different operation scenarios of the investigated industrial system.
In a possible embodiment of the apparatus according to the present invention, the apparatus further comprises a database storing the component library comprising component models of components and adapted to store the system model of the investigated industrial system generated by said generation unit.
In a further possible embodiment of the apparatus according to the present invention, the apparatus further comprises a control unit adapted to control at least one component of the investigated industrial system in response to the generated FMEA results.
The following provides according to the present invention an industrial system comprising hardware and/or software components and an apparatus for a model-based failure analysis of the complex industrial system consisting of said hardware and/or software components each represented by a context independent component model comprising interface terminals and a set of component behaviour modes including a normal mode and failure modes of the respective component stated as constraints on deviations, said apparatus comprising:
a generation unit adapted to generate a system model of the industrial system by loading component models of the components of the industrial system from a component library and connecting the interface terminals of the loaded component models according to a structure of the industrial system, and
a reasoning engine adapted to execute a constraint-based predictive algorithm to generate FMEA results for different operation scenarios of the industrial system.
Some of the embodiments will be described in detail, with references to the following figures, wherein like designations denote like members, wherein:
In the shown embodiment of
The apparatus 1 further comprises a reasoning engine 3 which is adapted to execute a constraint-based predictive algorithm to generate FMEA results for different operation scenarios of the investigated industrial system 7. In a possible embodiment, the generated FMEA results are used to predict a failure impact of a failure of one or several components on the functionality of the investigated industrial system 7. In a possible embodiment, the constraint-based predictive algorithm is executed by the reasoning engine 3 offline during design, maintenance and/or repair of the investigated industrial system 7. In a further possible embodiment, the constraint-based predictive algorithm is executed on the reasoning engine 3 online during operation of the investigated industrial system. The constraint-based predictive algorithm iterates over a Cartesian product of predefined operation scenarios OS and failure modes FM of each component or part to determine whether the failure propagation entails a local and/or system level effect E capturing a violation of a functionality of the investigated industrial system 7.
The database 4 comprises a component library of component models. Each hardware and/or software component is represented by a context independent component model CM comprising interface terminals and a set of component behaviour modes. These behaviour modes include a normal or okay mode and failure modes FM of the respective component. The different modes are stated in a preferred embodiment as constraints on deviations. The interface terminals of the component model are formed by channels to other components comprising interface variables exchanged with the other components of the investigated industrial system. In a possible embodiment, the component model CM of a component stored within the component library can comprise state variables indicating a state of the respective component. The component model further comprises a base model BM capturing a physical behaviour of the respective component. For instance, the base model BM can describe a physical and/or thermodynamic behaviour of the industrial system. In a possible embodiment, the component model CM comprises deviation models DM capturing deviations of actual values of variables from reference values of the respective variables. In a possible embodiment, the component model CM comprises also local effects indicating effects of component faults of the component on a functionality of the investigated industrial system 7.
In a further step S2, a constraint-based predictive algorithm is executed on a reasoning engine 3 to generate qualitative FMEA results FMEA-RES for different operation OS scenarios of the investigated industrial system 7.
The component model CM of a component 6 defines the behaviour of the component 6 and indicates the interaction of the component 6 with other components 6. The component model CM comprises interface terminals which represent channels to other components. The interface terminals comprise interface variables whose values are influenced by other connected components 6. For example, the interface terminal “output pressure” of one component is received by another component terminal as “input pressure”. For each component 6, one or more interfaces can be defined together with their types to allow exchange of information or data with other components. The interfaces are kept generic to allow changes. The connections are formed by links between two terminals of different components. When connecting terminals their types and variables match each other. In a possible embodiment, the component model CM of a component 6 does comprise interface terminals, state variables and parameters. Further, the component model CM comprises in a possible embodiment at least one base model BM, deviation models DM and local effects E for the respective component 6. A component 6 corresponds to an entity of the investigated industrial system 7. Each component or part can be an elementary component or an aggregation of other components. The component can be represented as classes in a hierarchy where components can inherit properties from parent components or superclasses. In a preferred embodiment, each component 6 is described with general conventions like a relation between a specific design and their direction of rotation. The component model CM comprises a set of component behaviour modes BM including one normal operation mode or okay mode NM and several possible failure modes FM. For example, considering an engine, the failure modes FM can comprise a higher torque and a lower torque of the engine. Further, the component model CM of a component 6 comprises a base model BM which forms the basis for different model variants. The constraint-based predictive algorithm executed in step S2 provides qualitative FMEA results. With the method according to embodiments of the present invention as illustrated in
The following table (Table 1) illustrates exemplary FMEA results provided by the method according to embodiments of the present invention for an exemplary industrial system formed by a core turbine engine such as illustrated by the physical model of
The components 6 of the investigated industrial system 7 must comply as much as possible with the physical system. After a component 6 has been identified, a corresponding component model CM can be loaded from the component library CL stored in the database 4. If a component model CM for the respective component 6 does not yet exist, a corresponding component model can be generated by a user or expert and stored in the component library CL. Component models CM are kept in preferred embodiment as generic as possible, i.e. context-free, so that the component model CM can be used for different systems (reusability). For example, the component model of an electric motor can be used in a loop or a system as well as in a core engine system, because its inherent functionality remains the same. The component model CM comprises one or several deviation models DM capturing deviations of actual values of variables from reference values of the respective variables. Qualitative deviation models DM are provided to determine potential failure causes and their effects. In the normal or okay behaviour mode NM of the component 6, the deviation of a variable is zero. In contrast, in a failure mode FM, the deviation is either positive or negative. The deviation can be expressed as Δx=xact−xref.
If all component models CM of all components 6 of the respective system 7 are available, they can be connected by means of an editor according to the topology of the investigated system 7. This means that one industrial system 7 can be configured or reconfigured using different topologies or structures STRU to provide different system models SM. After a specific system model SM of the investigated system 7 has been specified or selected, operation conditions or operation scenarios OS can be defined as input data. These operation scenarios OS can be stated as qualitative constraints on deviations. After having generated the system model SM in step S1, a constraint-based predictive algorithm can be run for a FMEA task. This constraint-based predictive algorithm is adapted to solve a finite constraint satisfaction problem FCSP which can be defined by a tuple (V,C,R), where:
V is a set of variables V={V1, V2, . . . , Vn} of the investigated industrial system with the domain DOM({Vi}). The domain can consist of a finite set of numbers or symbols and the variables of the system can have different domains. The overall domain is defined as a Cartesian product of the specific domains for each variable which defines the space in which the component behaviour can be specified:
DOM({Vi})=DOM(V1)×DOM(V2)× . . . ×DOM(Vn).
D is a function which maps the variables Vi to the domain DOM({Vi}).
R is a constraint which defines over a set of variables {Vi} in the domain DOM({Vi}) and characterizes a component, subsystem or system as RDOM({Vi}). A relation R is a constraint and substep of the possible behaviour space. The relation R contains elements which form a tuple. If the relation R is defined on a set of ordered variables, the set can be called a scheme of R and defined as scheme (R). The model fragments mentioned as Rij can be related to a behaviour mode Ei(cj) of the component cj. A mode assignment MA denotes the aggregated system of several modes of components 6 and specifies a unique behaviour mode for each of these components MA={mode Ei(cj)}.
The operation scenarios OS and failure modes FM are represented as a set of constraints or first order formulas. The constraint-based predictive algorithm iterates over the Cartesian product of the operation scenarios OS and failure modes FM and checks, whether they entail the defined failure mode via a constraint solver. It checks whether a given operation scenario OS and failure mode FM entails a local level and/or system level effect E or not. Effects E can also be stated as constraints and capture the violation of certain functionality. The FMEA results can be used to predict the failure impact on the functionality of the investigated system 7 in order to assess, whether they can lead to a critical situation where safety reliability requirements are violated. Further, the FMEA results can be used to minimize or mitigate any negative impact through a design correction of a system or a component design or through maintenance of the investigated system.
The compressed air from the compressor 6-2 enters a diffuser 6-6 which only propagates the airflow to the next component which is formed by the combustor. The air is heated up in the combustion chamber component 6-7. A burner 6-8 and a flame detection system 6-9 form part of the combustor section. The burner component 6-8 is used to mix the gas fuel with the compressed air in the combustion 6-7 and maintains stability of the flame. A gas fuel system 6-10 provides the required fuel to the burner 6-8 and the flame detection system 6-9 monitors the pilot and main flame during a start-up and operation phase.
Finally, the hot gas from the combustion chambers 6-7 enters the turbine 6-11. The turbine component 6-11 expands the air and drives the compressor 6-2 and a generator 6-12. A gearbox 6-13 transmits power from the turbine 6-11 to the generator 6-12. Ultimately, the generator 6-12 is operated to generate electricity for a power grid and the hot gas can be exhausted as exhaust air EA by a diffuser 6-14 to an air exhaust system 6-15.
A rotor assembly 6-16 illustrated in
Based on the sensor values provided by pressure and temperature sensors, an electronic control unit can generate commands to control the mechanical components of the investigated industrial system 7. The mechanical components can be controlled by specialized electronic control units ECUs 6-20. With the method and apparatus according to embodiments of the present invention, it is possible to perform a model-based failure analysis of a complex industrial system 7 such as the core gas turbine engine illustrated in
Table 1 illustrates the model-based generation of FMEA results for the core turbine engine. The start-up operation scenario happens when the motor is commanded to start to drive the compressor, air from inlet system is captured, valves take up their positions and rotation begins. During the start-up operation scenario, the motor, VGV, bleed valves positions are important and can affect the turbine and compressor. The operation scenario is reached when the turbine produces active power, the main flame is on and the rotor attains its maximum speed.
For the exemplary use case illustrated in
Domain, Terminals, Constants
Further, it is possible to define different terminals as illustrated in the following Tables 3 and 4:
For the different components, models can be defined in a specific embodiment as follows (Table 5):
These constraints can comprise the constraints listed in the following Table 6:
Although the invention has been illustrated and described in greater detail with reference to the preferred exemplary embodiment, the invention is not limited to the examples disclosed, and further variations can be inferred by a person skilled in the art, without departing from the scope of protection of the invention.
For the sake of clarity, it is to be understood that the use of “a” or “an” throughout this application does not exclude a plurality, and “comprising” does not exclude other steps or elements.
| Number | Date | Country | Kind |
|---|---|---|---|
| 15171927.5 | Jun 2015 | EP | regional |
This application claims priority to PCT Application No. PCT/EP2015/065842, having a filing date of Jul. 10, 2015, based on European Application No. 15171927.5, having a filing date of Jun. 12, 2015, the entire contents both of which are hereby incorporated by reference.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/EP2015/065842 | 7/10/2015 | WO | 00 |
