METHOD AND APPARATUS FOR THE MODEL-BASED MONITORING OF A TURBOMACHINE

Abstract

The method for the model-based monitoring of a turbomachine by means of a monitoring unit which receives relevant process parameters of the turbomachine, carries out calculations on the basis of these parameters and, as a result, determines one or more state values, by means of which the state of the turbomachine can be assessed, the calculations of the state values being carried out on the basis of one or more physical models of at least one subsystem of the turbomachine, wherein the at least one subsystem is an axial bearing of the turbomachine, and the one or more state values are selected from the axial thrust force, the axial bearing load and the axial shaft attitude.

Description

The present application incorporates the provisional U.S. Application 61/421,232 filed Dec. 9, 2010 by reference.

The present invention relates to a method for the model-based monitoring of a turbomachine by means of a monitoring unit which receives relevant process parameters of the turbomachine, carries out calculations on the basis of these parameters and, as a result, determines one or more state values, by means of which the state of the turbomachine can be assessed, the calculations of the state values being carried out on the basis of one or more physical models of at least one subsystem of the turbomachine. Furthermore, the invention relates to an apparatus for the model-based monitoring of a turbomachine.

Turbomachines are fluid energy machines which, as prime movers, convert various types of energy, for example thermal energy into mechanical energy or, as working machines, convert energy supplied into other energy states. Turbomachines have many different uses, for example in the form of steam turbines, gas turbines, water turbines or turbocompressors.

Turbomachines are usually operated continuously for several years and are shut down only for maintenance purposes. During this continuous operation, the function of components of the machine may be impaired, for example by wear, deposits or component failure. This may lead to a lowering of the efficiency of the machine, amounting to its complete incapacity to function. So that processes of this kind can be detected at an early stage and, optionally, measures can be taken to counteract the damage, such as wear, erosion or deposits, various monitoring and diagnostic methods are known in the prior art. Thus, conventionally, for safety reasons, limit values for individual important measurement variables are monitored, for example for bearing temperatures, oil pressures or the axial distances of the rotor from the casing. With the aid of oil analysis, information on abrasive particles in the lubricating oil and barrier oil and also on the change in oil quality can be obtained. Efficiency monitoring enables conclusions to be drawn as to effects, such as fouling or misflows. Oscillation diagnostics are used to detect imbalances and other causes of periodic mechanical oscillations.

In the abovementioned methods, the information to be obtained is based predominantly directly on the measurement of specific variables in or on the turbomachine. In addition, methods for the diagnosis and analysis of turbomachines are also known, which make use of models in order to obtain information.

The US patent application US 2007/0067114 A1 discloses a method for detecting wear phenomena on a turbine. In said method, different wear states are defined, for example the build-up of deposits or damage owing to particles, and a physically motivated model, for example a Kalman filter, is provided for each wear state.

Patent application EP 2 053 475 A1 describes a method for analyzing the operation of a gas turbine. In this case, a neuronal network is supplied with measurement values from the gas turbine, in particular with a dynamic pressure signal, and from these determines characteristic values which give information on normal operation and deviations from this.

Neuronal networks are an example of empirical black box models, the evidential force of which depends most critically upon the data on the basis of which they were set up. In the case of neuronal networks, historical data of selected measurement variables and of input variables and also useful output variables are used in order to generate links in the network. These links form the model, on the basis of which current output variables are calculated from current input variables after the training phase. The disadvantage of this, however, is that, on the one hand, a large quantity of training data is required in order to obtain a reliable model, and on the other hand, the evidential force of the models is restricted to the trained effects. A neuronal network is what is known as a black box model, which in this case means that the model does not enable conclusions to be drawn as to internal relationships in the fundamental real system, but only gives evidence of the input and output variables considered. This lack of transparency is a further disadvantage, since diagnostic function requirements can therefore be fulfilled only to a limited extent.

The object, therefore, is to provide an improved method for the monitoring of turbomachines, which, in particular, makes it possible to have evidence of effects which have not yet occurred in the past and which is transferrable at little outlay to other turbomachines.

According to the invention, this object is achieved by means of a method according to claim 1 and an apparatus according to claim 7. Preferred refinements of the invention are specified in the dependent claims 2 to 6 and 8 to 10.

According to the invention, a monitoring unit receives relevant process parameters of the turbomachine. Under consideration as process parameters are all data which can be detected by measurement on or in parts of the turbomachine or its auxiliary systems, for example the oil system. Furthermore, process parameters of the turbomachine may be data which can be detected by measurement on or in other technical systems connected to the turbomachine or its auxiliary systems. Examples of such process parameters are the properties, such as pressure, temperature, composition, flow rate or volume flow of the gas stream in a process gas feed line to a turbomachine or the power and rotation speed of an electric motor or of a steam turbine which drives a turbocompressor. The relevance of the process parameters depends on the subsystem considered and must be fixed in the individual case. Subsystems with their associated relevant process parameters are given by way of example further below. The detection of the process parameters and their transfer to the monitoring unit may take place in various ways, for example by electrical signals from a suitable measurement value transducer to the monitoring unit. Corresponding measurement and information methods and means are known to a person skilled in the art.

On the basis of the process parameters received, the monitoring unit carries out calculations and, as a result of these, determines one or more state values, on the basis of which the state of the turbomachine can be assessed. According to the invention, the calculations are carried out on the basis of one or more physical models of at least one subsystem of the turbomachine.

A physical model is to be understood as meaning that active relations existing in reality between process parameters and state values are mapped in the model. The degree of detailing in this case depends inter alia upon the useful state values and the available computing time. Examples of models and of state values calculated from them are given further below in connection with the corresponding subsystems.

Turbomachines, particularly in the chemical industry, are often designed and instrumented for the special application. In this case, inter alia, the respective process gas properties, such as pressure, temperature range or chemical composition, and also operational requirements, such as running capacity, redundant set-up, start-up and run-down processes and safety are to be taken into account. Various monitoring requirements can be fulfilled by means of the physical models according to the invention as a function of the respective instrumentation of an actual turbomachine.

In a preferred embodiment of the method according to the invention, the calculations of the state values are carried out on the basis of a plurality of physical models which in each case represent a subsystem of the turbomachine. The models of the subsystems are preferably set up as models, so that an overall model can be selected from the individual models and assembled, as required. For example, calculated state values can be linked to other state values or measurement values via equations, inequalities or characteristic diagrams by modeling. Of course, the respective models of the subsystems themselves may also have a modular setup, so that, for overall modeling, a type of construction kit of models and submodules can be adopted. The advantage of this embodiment is that a monitoring unit according to the invention can be adapted flexibly and efficiently to the most diverse possible types of construction, process conditions and instrumentations of a turbomachine.

In a preferred refinement of the method according to the invention, on the basis of relevant process parameters state values are calculated which are not detected or are not detectable by measurement, but contain important information on the state of the turbomachine. Examples of such state values are mechanical forces, bearing load or the generation of heat in a subsystem of the turbomachine. The monitoring of the turbomachine may in this case take place by comparing calculated state values with their respective limit values, for example the bearing load calculated from relevant process parameters with the limit value, given by the bearing material properties, of the maximum permissible bearing load.

In a further preferred refinement of the method according to the invention, on the basis of the relevant process parameters, state values are calculated which are detected by measurement in or on the turbomachine or its auxiliary systems. In this case, the monitoring of the turbomachine may mean that calculated state values are compared with the measurement values corresponding in each case to them. This type of monitoring is also designated as ‘software redundancy’. The size of the difference between a calculated and a measured value can give an indication of the state and quality of the measuring instrument. The size of the difference may also point to an untypical or defective state of the turbomachine. By a plurality of state values being considered in an integrated way, optionally, defective states can be diagnosed and the respective causes can be established.

In a preferred embodiment, the physical models used in the monitoring unit are based on conservation equations for energy, mass, momentum and/or forces. Such models are valid within a wide value range of the process parameters used, such as pressures or temperatures, and therefore it becomes possible to have evidence on effects which have not yet occurred in the past. This constitutes an appreciable advantage, as compared with empirical black box models. What may be considered as a further advantage is that physical models of this type can be transferred at little outlay to other turbomachines.

The term ‘subsystem’ is to be understood primarily in modeling terms. A turbomachine is usually constructed from a plurality of physically delimitable parts. Examples of delimitable parts of this type are axial bearings, radial bearings, coupling, rotor, shaft, casing, sliding ring, floating ring, lubricating oil tank, lubricating oil pump and lubricating oil filter. A subsystem in the context of this invention may constitute a model of an equipment part of the turbomachine, but it may also comprise a plurality of equipment parts, for example axial and radial bearings, barrier oil system or lubricating oil system.

In a further preferred embodiment, the method according to the invention is carried out online during the operation of the turbomachine. Relevant process parameters are in this case determined continuously or discontinuously. Depending on the model used and on the relevant process parameters required for this some parameters may also be determined continuously, while other parameters are determined discontinuously.

In a refinement according to the invention, the method is applied to a turbocompressor. In a preferred refinement, this is a multistage turbocompressor. In a turbocompressor, a plurality of subsystems for which the invention can advantageously be employed can be identified. Especially preferably, the at least one subsystem is selected from the barrier oil seal, oil filter and/or bearing, in particular axial bearing some subsystems are explained in detail by way of example further below. Moreover, preferably, the one state value or the plurality of state values contain characteristic information on

• • the generation of heat in a barrier oil seal,
  • the pressure loss across an oil filter,
  • the axial shaft attitude in a bearing, and/or
  • the bearing load in a bearing, in particular an axial bearing.

Furthermore, the method according to the invention can advantageously be applied to a checking of protective devices of the turbomachine. With the aid of protective devices, critical values of a turbomachine, for example the axial shaft attitude, also called the axial distance, are usually monitored. In the example of axial distance monitoring, the position of the rotor in relation to the compressor casing is monitored by distance sensors, in order reliably to prevent the rotor from brushing against the casing. The measurement signals from the redundantly designed sensors must lie within a range defined by limit values when the machine is operating normally. If the limit value is overshot by one of the signals, the machine is usually driven automatically into a safe state. One advantage of the method according to the invention is that the axial distance to be expected can be calculated from process parameters, such as pressures, temperatures, flow rates and coupling forces, in the turbomachine and can be compared with the axial distance values detected by measurement. As a result, those time variations of the measurement signal which are generated by variations in the process parameters can be distinguished from those which occur due to damage, wear, deposits, faults in the sensor and evaluation electronics or other faults. Thus, by virtue of the method according to the invention, the early detection of machine damage can be markedly improved, as compared with the prior art. Axial distance monitoring is only one example of the checking of a protective device of a turbomachine on the basis of a calculated state value. Further state values suitable for checking are, for example, the barrier oil outlet temperature and the calculated power loss in the barrier oil seal, as described further below.

In a preferred embodiment, historical data are also resorted to, in addition to current data, in order to monitor one or more protective devices. Historical data are to be understood as being relevant process parameters, the values of which were detected in the past and were stored in a data storage system. The data storage system is preferably a storage system integrated in a process management system or is what is known as an operating data information system. An operating data information system is advantageously installed on a separate computer and has data interfaces with the sensors of the turbomachine or with a process management system which, in turn, is connected by information communication to the sensors of the turbomachine. Sensors are in this context understood as meaning the whole of all the measuring instruments on the turbomachine and its auxiliary systems, for example for the detection of temperature, pressure, mass flow, volume flow, distances, current or rotational speed.

The historical data contain information on the behavior of the turbomachine under normal operating conditions and, optionally, in a defective state. In combination with machine and process characteristics, the historical data allow the modeling of the normal machine behavior and the validation of the models. If the turbomachine already had damage before modeling, by means of the historical data the time development of the damage can be analyzed and the diagnostic algorithm can thereby be validated and, optionally, refined. The inclusion of historical data is also advantageous particularly when measuring instruments are to be monitored with regard to their time behavior, for example phenomena, such as the drift of measurement values.

In an advantageous refinement, the method according to the invention is implemented in the form of a program code for a computer program, which is suitable for carrying out the method when the computer program is executed on a suitable computer installation.

The invention comprises, furthermore, a computer program product with a computer-readable medium and with a computer program stored on the computer-readable medium and having program code means which are suitable for carrying out the method according to the invention when the computer program is run on a suitable computer installation.

The computer program is preferably prepared in a programming language familiar to a person skilled in the art and is adapted to the respective hardware and software requirements of the computer installation used. Especially preferably, the computer program has a modular setup. This makes it possible to have an efficient implementation of monitoring algorithms for the diverse monitoring requirements, the machine designs and equipment variants of measurement technology.

A further aspect of the invention is an apparatus for the model-based monitoring of a turbomachine. The apparatus comprises a monitoring unit with a calculation unit and with an assessment unit, the monitoring unit being prepared for receiving relevant process parameters from the turbomachine, calculations being capable of being carried out in the calculation unit on the basis of these parameters and, as a result, one or more state values being capable of being determined, and the state of the turbomachine being capable of being assessed in the assessment unit on the basis of the state values, and the monitoring unit being set up in such a way that the calculations can be carried out on the basis of one or more physical models of at least one subsystem of the turbomachine.

Preferably all the calculation instructions, such as equations, inequalities and/or characteristic diagrams of the at least one physical model, which are necessary so that state values can be calculated on the basis of the relevant process parameters, are stored in the calculation unit.

The calculated state values are used in the assessment unit in order to assess the state of the turbomachine. As a result of the assessment, for example, a report can be prepared for the maintenance personnel, for example at fixed time points or triggered by results of the assessment unit, such as the overshooting of a limit value. A report of this type may contain both information on the current time point and information on the past or the future. An example of past data is the textual and/or graphical illustration of the development of calculated state values over a specific period of time in the past. An example of future data is prognostic values for state values over a specific period of time in the future which have been calculated on the basis of the one or more physical models.

In a preferred embodiment, the monitoring unit is implemented completely or partially on a computer installation, and the apparatus comprises, furthermore, means for transferring the relevant process parameters from the turbomachine to the monitoring unit.

Suitable computer installations are, for example, personal computers (PCs), industrial PCs, process management systems (PLS), stored-program controls (SPS), safety SPS (SSPS), microprocessors embedded into the turbomachine or else central data processing platforms, such as manufacturing execution systems. Those means suitable for transferring the relevant process parameters are, for example, wired signal transmission systems, such as the 4-20 mA signal customary in the process industry, or field bus systems, but also wireless systems, such as radio links. The means for transferring relevant process parameters may be implemented as individual connections or in the form of networks, for example as local area networks (LAN), wireless LAN, Internet or Intranet connections.

In a further-preferred refinement, the means comprise a system for the storage and processing of the relevant process parameters. This system is especially preferably an operating data information system.

In a further preferred embodiment, the monitoring unit is implemented completely or partially in the form of function blocks in a process management system.

Different actions can be triggered by the monitoring unit as a function of the calculated state values and of the fundamental assessment criteria. Thus, for example, a warning message or alarm message can be sent or automatic intervention in process management can be carried out. The monitoring unit may also be configured in such a way that, in the presence of specific preconditions, the turbomachine is shut down automatically, for example when a critical situation is established on the basis of the calculated state values and assessment criteria. The actions may also be executed so as to be staggered in time or hierarchically.

The method according to the invention makes it possible to detect at an early stage faulty states or wear states which it has not been possible to detect hitherto. Examples of these are a high axial force load upon the axial bearings, skewing of the bearing, carbon build-up on a floating ring or the tilting of a floating ring. It is possible, furthermore, to calculate the time development of state values on the basis of the models over a certain period of time into the future. On the basis of the information obtained with the aid of the models, measures can be taken to prevent damage, thus leading to a reduction in failures, a reduction in maintenance costs and often also a lengthening of the running time of the turbomachine. As a result, maintenance intervals can be prolonged and shutdown times can be minimized, thus entailing sometimes considerable cost benefits for the operators of the turbomachine. Also, independently of a prolongation of the operating time, it is in the interests of the operator to obtain information on the state of the turbomachine, for example in order to optimize operation or to increase the rate of utilization. The method according to the invention, because of its flexibility, can be applied to the most diverse possible turbomachines. It allows continuous operation-accompanying monitoring of the turbomachine and its auxiliary systems, increases safety and reduces the risk for personnel and the environment.

The invention is explained in more detail below by means of the drawings, which are to be understood as basic illustrations. They do not constitute any restriction of the invention, for example with regard to actual design variants. In the drawings:

FIG. 1: shows a basic diagram of a turbomachine with a monitoring unit,

FIG. 2: shows a basic diagram of a turbomachine with an alternative implementation of a monitoring unit,

FIG. 3: shows submodules and their interlinking,

FIG. 4: shows an illustration of the monitoring of axial distance measurement on a turbocompressor.

FIG. 1 shows diagrammatically the coupling of a monitoring unit 20 to a system 10 to be monitored, which contains the turbomachine, its auxiliary systems and other systems which generate measurement for the monitoring task. To simplify the description, this system to be monitored is designated below as the ‘turbomachine 10’. The monitoring unit 20 has a calculation unit 22 in which calculations are carried out. The models on which the calculations are based may comprise a plurality of submodels 24. Furthermore, the monitoring unit 20 comprises an assessment unit 26 in which assessments are carried out. The turbomachine 10 comprises a plurality of subsystems 12, and in this case these may be physically delimitable subsystems of the machine, for example axial and radial bearings, barrier oil system or lubricating oil system. Submodels 24 on which the calculation is based may be models of respective subsystems 12 of the machine. However, a subsystem 12 may also be mapped with the aid of a plurality of submodels 24, and therefore the number of submodels 24 may be markedly larger than the number of subsystems 12.

The monitoring unit 20 receives relevant process parameters 14 from the turbomachine 10. In this case, the parameters 14 may be transferred directly, as indicated in the right hand part of the diagram. The parameters 14 may also initially be supplied to a system 30 for the storage and processing of the relevant process parameters, in particular to an operating data information system. Processed parameters 16 are transferred from this system to the monitoring unit 20.

State values 28, on the basis of which the state of the turbomachine can be assessed in the assessment unit 26, are determined from the relevant process parameters 14, 16 in the monitoring unit 20. In order to make state values 28 and assessment results 29 visible to the operator, they are transferred to an indicator device 40, for example a display of a data processing device.

FIG. 2 shows an alternative form of implementation of the method according to the invention. In this case, the submodels 24 of the calculation unit 22 and the assessment unit 26 are implemented as function blocks in a process management system 50. The system 30 for the storage and processing of the relevant process parameters and also the indicator device 40 are also part of the process management system 50. Suitable process management systems and the information coupling of these to a turbomachine are known to a person skilled in the art.

Example

Various submodels 24 were prepared for a turbocompressor and implemented on a central manufacturing execution system platform which affords the possibility for the modular implementation of complex algorithms. FIG. 3 gives a diagram of the individual submodels, the numbers signifying the following:

• • 240 . . . Thermodynamics model
  • 241 . . . Orientation model
  • 242 . . . Shaft sealing model
  • 243 . . . Coupling model
  • 244 . . . Thrust model
  • 245 . . . Rotor model
  • 246 . . . Model of the auxiliary system barrier oil
  • 247 . . . Model of the auxiliary system lubricating oil
  • 248 . . . Axial and radial bearing model
  • 249 . . . Casing model

The arrows indicate that results from the respective submodel are incorporated into another submodel.

The oscillation model which receives data from oscillation sensors, for example frequency spectra of the oscillations, as part of the relevant process parameters is not depicted. This model has access to the orientation model, the shaft sealing model, the coupling model, the rotor model, the axial and radial bearing model and the casing model.

The invention is explained in more detail below by means of some submodels. What all of these have in common is that they are physical models of subsystems of the turbomachine. The actual approaches to modeling, particularly the degree of detailing, can be selected and adapted as a function of the monitoring requirements. The submodels illustrated below are to be considered merely as examples. They do not signify any restriction of the invention to the actual relevant process parameters, modeling approaches and state values determined from these.

In FIG. 3, the barrier oil system, as it is known, is designated by reference numeral 246. The barrier oil system serves, inter alia, for degassing, cooling and purifying the barrier oil and for supplying it in a regulated manner to the shaft seals. The barrier oil system was subdivided by modeling into a plurality of submodels, the submodel barrier oil tank, the submodel oil filter, the submodel oil pump, the submodel oil cooler, the submodel barrier oil seal and the submodel valves and regulating valves.

The modeling of the submodels of the barrier oil tank, oil pumps, oil coolers and valves and regulating valves obeys the modeling approaches known to a person skilled in the art, such as the mass and energy balance, heat transmission law for oil coolers, characteristic curves for pumps and valves. The oil filter was mapped by the physical model described below. The oil filter serves for removing impurities in the oil. These impurities may be, for example, abrasive particles, carbon particles or precipitating additives. A significant increase in impurities indicates a problem in the barrier oil system, such as local excessive temperature rises of the oil or a lack of lubrication with metal abrasion. The oil filter model is advantageously suitable for monitoring the contamination state and, in the event of significant variations, for generating a warning message.

Relevant process parameters in the context of the present invention are, for the submodel oil filter, the following:

• • the pressure difference Δp across the oil filter,
  • the temperature T of the oil, measured upstream or downstream of the oil filter or an average value from these, and
  • the volume flow V of the oil through the filter. Alternatively for the measured volume flow V, a calculating value could also be used by dividing the measured oil mass flow by the oil density.

These parameters were detected online by means of known measuring instruments. Furthermore, the temperature-dependent viscosity of the oil η(T) was required for modeling the oil filter.

On the basis of these parameters, a specific filter parameter k as a state value, which cannot be detected by measurement, was determined with the aid of the model

$k=V\frac{\eta \left(T\right)}{\Delta p}$

based on Hagen's and Poiseulle's law for laminar-throughflow filters. If the value of the filter parameter decreases over time, this is an indicator of increased filter resistance, for example because of particle deposits. The present state and the time duration until the necessary exchange of the filter can therefore be judged by the assessment of the filter parameter determined by the model. A warning message is triggered when the value of the filter parameter or the rate of change in time of the filter parameter overshoots the limit value. The warning message may point to increased wear and particle deposition or to a rapid change in the oil quality, for example due to cokling. Thus, important components, in this case the oil filter and oil quality, can be monitored with the aid of the method according to the invention for the reliable operation of the turbomachine.

The submodel of the barrier oil seal comprises, as a balance space, the oil distribution ducts in the turbomachine and the sealing element itself, in this example a combined floating-ring/sliding-ring seal.

The following are selected as relevant process parameters for the barrier oil seal:

• • the inlet temperature of the barrier oil,
  • the outlet temperature of the barrier oil,
  • the differential pressure of the barrier oil across the seal,
  • the casing temperature,
  • the rotational speed of the rotor, and
  • the mass flow of the barrier oil.

These process parameters were detected online by means of known measuring instruments and methods. In order to determine the mass flow of the barrier oil, the volume flow of the barrier oil may also be measured and used as a process parameter. In this case, the mass flow can be calculated from the volume flow and the temperature-dependent density of the oil. Furthermore, the temperature-dependent specific heat capacity of the oil and the temperature-dependent viscosity of the oil were required in order to model the barrier oil seal. These two variables were taken from characteristic diagrams for the oil used.

The fundamental modeling approaches are a mass balance and an energy balance which take into account the following introductions of energy: the frictional power in the oil gap of the sealing element of the floating ring, the internal frictional power of the oil through contact with rotating surfaces and dissipation, the introduction of heat via the casing wall of the oil ducts, the introduction of heat via the rotor shaft and the introduction of heat from the sliding ring seal. The respective energy fractions were modeled as a function of a plurality of process parameters and adapted to historical measurement data from the normal operation of the turbomachine.

The power loss in the barrier oil seal is calculated from

P_loss=P_V+P_GR+P_SR

P_V=c_Vn²η_dyn(θ_m,V)

P_GR=c_GRμ_frictnΔp_GR

P_SR=c_SRn²η_dyn(θ_m,SR)

in which P_loss is the overall power loss, P_V, P_GR, P_SR are respectively the ventilation, sliding ring and filtering ring power loss, n is the rotational speed of the shaft, Δp_GR, is the pressure difference across the sliding ring, θ_m,V, θ_m,GR, θ_m,SR are the average oil temperatures respectively in the ventilation region, of the sliding ring and of the filtering ring, η_dyn is the temperature-dependent dynamic viscosity of the barrier oil, c_V, c_GR, c_SR are constants dependent on the geometry of the seal (such as, for example, the average ventilation, sliding ring and floating ring diameters, the length of the ventilation gap, width of the ventilation gap, floating ring width), μ_frict is the coefficient of friction between the sliding ring and the rotor.

The heat quantity discharged by the barrier oil may be calculated as follows:

P_oil=P_oil,V+P_oil,GR+P_oil,SR

P_oil,V=2{dot over (m)}c_p(θ_m,V)(θ_out,V−θ_m,V)

P_oil,GR=2{dot over (m)}c_p(θ_m,GR)(θ_out,GR−θ_m,GR)

P_oil,SR=2{dot over (m)}c_p(θ_m,SR)(θ_out,SR−θ_m,SR)

in which P_oil is the overall heat discharged by the oil, P_oil,V, P_oil,GR, P_oil,SR are in each case the heat discharged by the oil from the ventilation part, sliding ring and floating ring, {dot over (m)} is the mass flow of the barrier oil, θ_out,V, θ_out,GR, θ_out,SR are the temperatures of the barrier oil in each case at the outlet point from ventilation, the sliding ring and the floating ring, and c_p is the temperature-dependent specific heat capacity of the barrier oil.

The energy balance results in:

P_V=P_out,V,P_GR=P_out,GR,P_SR=P_out,SR

θ_in,V=θ_in,θ_out,V=θ_in,GR,θ_out,GR=θ_in,SR,θ_out,SR=θ_out

in which θ_in is the inlet temperature of the barrier oil, θ_out is the outlet temperature of the barrier oil and θ_in,V, θ_in,GR, θ_{in, SR} are the temperatures of the barrier oil in each case at the inlet point into the ventilation, sliding ring and floating ring.

On the basis of the relevant process parameters, the following state values were calculated with the aid of the model described above:

• • the temperature of the barrier oil at the inlet point (without the use of the measured inlet temperature),

the temperature of the barrier oil at the outlet point (without the use of the measured outlet temperature),

• • the temperatures of the barrier oil at a plurality of selected points within the barrier oil seal,
  • the power loss of the barrier oil seal, and
  • the heat quantity discharged by the barrier oil.

In general, the temperatures of the barrier oil should not be too high. If the temperature is too high, the oil would be unstable, and soiling in the gap and failure of the seal would occur. The power loss is important information on the heat occurring in the turbomachine during operation. In order to judge the instantaneous state of the barrier oil seal and predict its future behavior, the state values are assessed as follows.

Since the inlet temperature of the barrier oil was both calculated as a state value and measured as a relevant process parameter, it is a question here of software redundancy as described above. By the difference between the calculated and the measured inlet temperature being considered, it was possible to monitor the functioning capacity of the temperature sensor and problems possibly occurring in the barrier oil seal. An alarm is triggered when the difference between the calculated and the measured inlet temperature overshoots a limit value. The state value of the outlet temperature of the barrier oil was used in a similar way for monitoring.

The temperatures of the barrier oil within the barrier oil seal are mostly not detected by measurement. The calculated values of these temperatures were compared with their respective limit values. By one signal being triggered when the limit value is overshot, the service and maintenance personnel can be informed in due time of high temperatures within the barrier oil seal.

Furthermore, the state values of the power loss and of the heat quantity were also compared with their corresponding limit values. For example, a rising power loss may point to increasing friction within the seal, and this may be attributable to an increasing deposition of carbon in the seal. With the time profile of the power loss being taken into account, in this example the optimal time point for cleaning the seal can be determined. With the aid of this information, the operation can likewise be adapted in such a way that the cleaning time point is delayed.

A further example is the subsystem axial bearing. The axial bearing holds the rotor in its position in relation to the casing by absorbing the axial rotor forces. There are various forms of construction for axial bearings, for example tilting segment axial bearings. In all cases, a thin oil film which is maintained via hydrodynamic forces is located between the two bearing surfaces.

The following relevant process parameters were fixed for the submodel axial bearing:

• • the suction pressure and suction temperature of the gas stream to be compressed,
  • the end pressure and end temperature of the compressed gas stream,
  • the gas composition,
  • the intake quantity of the gas as the mass flow,
  • the stage pressure and stage temperature upstream and downstream of each individual wheel,
  • the rotational speed of the rotor,
  • the axial bearing temperatures, also on individual bearing blocks of the multipart segment bearing, and
  • the axial distance of a reference point of the rotor in relation to the casing.

All the relevant process parameters were detected online by means of known measuring instruments and methods. In order to determine the mass flow of the gas, the volume flow of the gas may also be measured and used as a process parameter. The mass flow is then calculated from the product of the volume flow and specific density of the gas stream.

Essentially two modeling principles were adopted in preparing the model for the axial bearing. On the one hand, thermodynamically based models were set up for determining the axial forces, axial bearing load, compressor power and efficiency from the relevant process parameters. To determine the axial distance, physically motivated correlations were set up on the basis of historical data of relevant process parameters, such as axial bearing temperatures, mass flow or volume flow, rotational speed or axial thrust force, in this case a modified Hookes' spring law which models the axial displacement of the rotor in relation to the casing as a function of the calculated axial forces. This model was extended by modeling approaches for taking into account the influence of further process parameters, such as temperature influences upon the mechanical rigidity of the system and the oil film thickness in the bearing gap. FIG. 4 illustrates the monitoring of the axial distance by model estimation. The abscissa axis represents the time in month/day format, the axial distance being plotted on the ordinate axis. When the measurement value leaves the confidence interval, as was the case, for example, on September 20 and on October 2, this gives an indicator of a machine state which does not correspond to the modeled normal behavior of the machine.

An exemplary model is the following energy balance for the axial bearing. The frictional heat {dot over (Q)} occurring in the axial bearing can be calculated, on the one hand, from the heat flow discharged by the lubricating oil

{dot over (Q)}=ΔT·c_p·{dot over (m)}  (Equation 1),

ΔT designating the temperature rise during the flow through the bearing, {dot over (m)} the oil quantity flow and c_p the heat capacity. However, the frictional heat {dot over (Q)} can also be calculated from the frictional power by

{dot over (Q)}=F·μ·π·d_m·n  (Equation 2),

F designating the thrust force acting on the bearing (calculated from the thrust model), μ designating the coefficient of friction, d_n, the average bearing diameter and n the rotational speed. With the exception of μ and {dot over (Q)}, all the variables in equations 1 and 2 are measurable. Consequently, μ and {dot over (Q)} can be calculated from these two equations and used for monitoring the friction state in the axial bearing.

With the aid of the models, the following state values for the axial bearing were determined:

• • the axial thrust force including momentum, pressure force, piston force and frictional force,
  • the axial bearing load as force per unit area,
  • the axial distance of a reference point of the rotor in relation to the casing, and furthermore,
  • the compressor power and the efficiency which relate to the entire turbomachine.

The state values are used in various ways for the diagnosis as follows.

Turbomachines are often equipped with a plurality of axial distance probes. The axial distance probes are essential for machine safety, and therefore alarm and shutdown limits are placed on their values. Experience shows, however, that the axial distance probes often indicate different values even shortly after commissioning. Determining the axial distance by modeling on the basis of relevant process parameters measured online affords a possibility for checking the measurement values of the axial distance probes in the form of software redundancy. By comparison of the measurement values and of the axial distance determined by the model, damage to the axial distance probes or to the turbomachine can be detected at an early stage.

By means of the state values, such as the axial thrust force and axial bearing load, insights into the load state of the axial bearing of the turbomachine were obtained, which have not hitherto been available in this form online during continuous operation. Thus, for example, alarm messages can now be generated as soon as the axial thrust force or axial bearing load reach predetermined limit values.

Further important information which can be obtained from the state values of the models according to the invention are evidence as to the efficiency of the turbomachine. Thus, for example, the compressor power, the efficiency or its time change can be put into relation with relevant process parameters, such as the gas stream. Consideration of the time profiles of such relations enables conclusions to be drawn as to phenomena, such as fouling or erosion in the turbomachine, which may be reflected, for example, in a variation in surface roughness or in the actually usable geometries.

The above-explained submodels oil filter, barrier oil seal and axial bearing are independent of one another and may both be integrated jointly into an overall model and used separately. If, for example, only the barrier oil system is to be monitored, the submodel axial bearing is not required for this purpose. Depending on the monitoring requirements of a turbomachine, a corresponding model may be set up. Examples of further submodels which may be used advantageously within the scope of the invention are listed in the following tables.

Thermodynamics model of the process gas compressor
Relevant process parameters      State values
Machine geometry (for example, number of stages, Power;
wheel diameter);                 Flow velocities;
Gas inlet temperature and pressure; Temperatures;
Gas outlet temperature and pressure; Pressures;
Stage inlet temperature and pressure; Efficiency;
Stage outlet temperature and pressure; Fouling
Rotational speed;
Gas composition;
Intake quantity flow (volume flow or mass flow);
Setting pump protection regulating valve;
Prerotation setting (inlet guide vane setting);
Pressure, temperatures and flow rates at intermediate
coolers and liquid separators;
Flow rate, composition and temperature of liquid
injections;
Characteristic diagrams

Thermodynamics model of the steam turbine
Relevant process parameters      State values
Machine geometry (for example, number of stages, Power;
wheel diameter);                 Flow velocities;
Fresh steam pressure, temperature and quantity; Temperatures;
Wheel space pressure; extraction pressures; Pressures;
Evaporation pressure and temperature; Efficiency;
Overflow pressure;               Fouling
Setting regulating valves, extraction valves,
overflow valves;
Rotational speed;
Characteristic diagrams

Orientation model
Relevant process parameters      State values
Geometry data of the machine set-up (inter alia, the Thermal growth;
position of the steam turbine in relation to the Support forces;
foundation and the position of the process gas com- Expansions
pressor in relation to the foundation);
Temperatures of the ambient air, of the foundation
and of the machine casing;
Calculated forces

Shaft sealing model
Relevant process parameters      State values
Geometry data;                   Heat capacity;
Power; rotational speed;         Fouling;
Temperatures of the casing and rotor; Oil temperature within
Barrier oil quantity, pressure and temperature; the seal for monitoring
Barrier oil substance properties (viscosity, the oil decomposition or
heat capacity)                   carbon formation
                                 reaction

Coupling model
Relevant process parameters      State values
Geometry data; shaft connection; Coefficient of friction;
Rotational speed; axial force; torque; Heat generation
Surface and lubrication oil temperature;
Lubricating oil substance properties (viscosity,
heat capacity);
Oscillation data; oil flow rate

Thrust model
Relevant process parameters      State values
Data of the thermodynamics model, of the axial and Forces;
radial bearing model and of the coupling model Power;
                                 Axial distance

Rotor model
Relevant process parameters      State values
Geometry data (for example, rotor dimensions, Forces and expansion
mounting, shaft connection);
Rotational speed; drive power;
Calculated or measured forces at the rotor;
Steam or gas pressures, temperatures and flow
velocities;
Imbalance; accelerations

Model of the auxiliary system barrier oil
Relevant process parameters      State values
Barrier oil pressure, temperature and flow Filter parameters;
rate upstream and downstream of the turbo- Heat generation;
machine and upstream and downstream of the Monitoring of the
oil coolers and filters;         functioning capacity of
Level and temperature in the barrier oil the sensors and
tank;                            actuators by utilization
Pump switchings (on/off) of the main oil of redundancy and
pump and auxiliary oil pump;     special operating states;
Cooling water temperature and inlet and Fouling
outlet of the oil coolers and cooling
water quantities;
Setting of oil temperature regulating
valve and pressure regulating valve;
Sealing pressure

Model of the auxiliary system lubricating oil
Relevant process parameters      State values
Lubricating oil pressure, temperature Filter parameters;
and flow rate upstream and downstream Heat generation;
of the turbomachine and upstream and Monitoring of the
downstream of the oil coolers and functioning capacity of
filters;                         the sensors and
Level and temperature in the lubri- actuators by utilization
cating oil tank;                 of redundancy and
Pump switchings (on/off) of the main special operating states;
oil pump and auxiliary oil pump; Fouling
Cooling water temperature and inlet and
outlet of the oil coolers and cooling
water quantities;
Setting oil temperature regulating
valve and pressure regulating valve

Axial and wheel bearing model
Relevant process parameters      State values
Geometry data and type of construction; Axial distance;
Oil pressure, temperature and viscosity; Bearing load (surface
Axial force; axial distance;     pressure)
Bearing temperature; bearing material

Casing model
Relevant process parameters      State values
Geometry data; component plays;  Support forces;
Calculated forces and expansions in the turbomachine Expansions

Oscillation model
Relevant process parameters      State values
Oscillation measurements (relative oscillation, Deviation of the
absolute oscillation, phase position, orbit, oscillation
frequencies);                    measurements from
Balancing data;                  process parameter-
Calculated forces and pressures in the dependent
turbomachine;                    characteristic diagram
Power;
Casing expansion

Model of the auxiliary system barrier gas
Relevant process parameters      State values
Barrier gas differential pressure and temperature, Filter parameter;
sealing pressure, filter differential pressures; Heat generation;
Barrier gas pressure             Monitoring of
(designation for gases entering the gas seal/the functioning capacity of
primary seal in the case of tandem seals) the sensors and
Buffer gas pressure              actuators by utilization
(designation for gases entering the seal for oil of redundancy and
separation)                      special operating
Scavenging gas pressure          states;
(designation for gases entering the seal for Fouling
process separation)
Torch gas pressure
(designation for leakage gases of the primary seal
of a tandem seal or of a tandem seal with inter-
mediate labyrinth)
Primary gas pressure
(designation for the barrier gas of the process-
side seal of a tandem seal or of a tandem seal
with intermediate labyrinth)
Secondary gas pressure
(nitrogen which is switched on the safety seal
of a tandem seal with intermediate labyrinth)
Barrier gas, buffer gas, scavenging gas, primary
gas and secondary gas quantity

Model of the regulating oil system
Relevant process parameters      State values
Regulating oil pressure;         Diagnosis of friction values,
Differential pressure regulating oil filter; slack and seizure on valves;
Start-up oil pressure for quick-action Monitoring of the function
closing valves;                  capacity of the sensors and
Quick-action closing oil pressure; actuators by utilization of
Test oil pressure for part stroke test quick- redundancy and special
action closing valve             operating states;
                                 Fouling

A further advantage of the modular model set-up is to be seen in the simple transferability of an already prepared model to another turbomachine. On the one hand, only the submodels required for the new model need to be selected. On the other hand, depending on the set-up of the turbomachine, identical or similar submodels with a low adaptation requirement can be adopted, and therefore new models have to be prepared or complicated adaptations carried out solely for markedly different subsystems. This procedure saves time and cost and increases the reliability of the models when proven and tested submodels can be resorted to.

Claims

1. A method for the model-based monitoring of a turbomachine by means of a monitoring unit which receives relevant process parameters of the turbomachine, carries out calculations on the basis of these parameters and, as a result, determines one or more state values, by means of which the state of the turbomachine can be assessed, the calculations of the state values being carried out on the basis of one or more physical models of at least one subsystem of the turbomachine, wherein the at least one subsystem is an axial bearing of the turbomachine, and the one or more state values are selected from the axial thrust force, the axial bearing load and the axial shaft attitude.

2. The method according to claim 1, wherein the physical models are based on conservation equations for energy, mass, momentum and/or forces.

3. The method according to claim 1 or 2, wherein the method is carried out online during the operation of the turbomachine on the basis of continuously and/or discontinuously determined relevant process parameters.

4. The method according to one of claims 1 to 3, wherein the turbomachine is a turbocompressor, in particular a multistage turbocompressor.

5. The method according to one of claims 1 to 4, wherein a protective device of the turbomachine is checked on the basis of at least one calculated state value.

6. The method according to one of claims 1 to 5, wherein historical relevant process parameters which are detected in the past and are stored in a data storage system are used additionally for the calculations.

7. An apparatus for the model-based monitoring of a turbomachine, comprising a monitoring unit with a calculation unit and with an assessment unit, the monitoring unit being prepared for receiving relevant process parameters from the turbomachine, calculations being capable of being carried out in the calculation unit on the basis of these parameters and, as a result, one or more state values being capable of being determined, and the state of the turbomachine being capable of being assessed in the assessment unit on the basis of the state values, the monitoring unit being set up in such a way that the calculations can be carried out on the basis of one or more physical models of at least one subsystem of the turbomachine, wherein the at least one subsystem is an axial bearing of the turbomachine.

8. The apparatus according to claim 7, wherein the monitoring unit is implemented completely or partially on a computer installation, and the apparatus comprises, furthermore, means for transferring the relevant process parameters from the turbomachine to the monitoring unit.

9. The apparatus according to claim 8, wherein the means comprise a system for the storage and processing of the relevant process parameters, in particular an operating data information system.

10. The apparatus according to claim 7, wherein the monitoring unit is implemented completely or partially in the form of function blocks in a process management system.