WO2015149928A2 - Method and device for online evaluation of a compressor - Google Patents

Method and device for online evaluation of a compressor Download PDF

Info

Publication number
WO2015149928A2
WO2015149928A2 PCT/EP2015/000675 EP2015000675W WO2015149928A2 WO 2015149928 A2 WO2015149928 A2 WO 2015149928A2 EP 2015000675 W EP2015000675 W EP 2015000675W WO 2015149928 A2 WO2015149928 A2 WO 2015149928A2
Authority
WO
WIPO (PCT)
Prior art keywords
compressor
model
undegraded
degradation
performance
Prior art date
Application number
PCT/EP2015/000675
Other languages
French (fr)
Other versions
WO2015149928A3 (en
Inventor
Matteo CICCIOTTI
Ala Eldin Farag BOUASWAIG
Stefan MANSS
Ricardo Martinez-Botas
Original Assignee
Basf Se
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Basf Se filed Critical Basf Se
Publication of WO2015149928A2 publication Critical patent/WO2015149928A2/en
Publication of WO2015149928A3 publication Critical patent/WO2015149928A3/en

Links

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B17/00—Systems involving the use of models or simulators of said systems
    • G05B17/02—Systems involving the use of models or simulators of said systems electric
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
    • G05B23/0243—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model
    • G05B23/0254—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks

Definitions

  • the present invention refers to a method for online evaluation of a compressor.
  • Compressors are used m a wide area such as for compressing fluids or gases to run, for example, a plant or an engine.
  • a compressor can be degraded by the very high frequency of rotations of a turbine within the compressor and the very high temperatures of the exhausting gases which may lead to a temperature gradient within the compressor and, therefore, causing strains within the material of e.g. an impeller and/or the compressor itself. Further, parts of a compressor are exposed to a particular environment and, therefore, prone to slow changes such as fouling, erosion or corrosion. Thus, compressors are under high mechanical stress which may lead to a decreased performance of the compressor and/or a respective combustion engine or plant .
  • Another approach known in the state of the art is the use of model-based pattern recognition approaches which are trained on a set of training data and programmed to detect specific patterns which may be correlated to a malfunction .
  • the method for online evaluation of an operative range and performance of a compressor comprises at least the following offline steps: a) setting up a digital undegraded model of the compressor in an undegraded state of the compressor; b) calibrating and validating the undegraded model offline using historical data from a compressor running line; c) calculating at least one undegraded performance map using the undegraded model; and at least the following online steps: d) calculating a degradation-adaptive model by updating the undegraded model with operative data of the compressor determined by at least one sensor of the compressor; e) calculating at least one actual performance map of an operative state of the compressor running line; f) detecting malfunctions of the compressor by comparing the at least one undegraded performance map derived by the undegraded model and the at least one actual performance map derived by the degradation- adaptive model via at least one mathematical function .
  • the terms "turb) setting up a digital undegraded model of the compressor in an undegraded state of the compressor comprises at least the following offline steps: a) setting
  • the described method is based on an undegraded model and a degradation-adaptive model, wherein the undegraded model is set up in a first step offline.
  • the undegraded model is set up in a non-operative state of a particular compressor, wherein the undegraded model is based on a physical modeling of the compressor.
  • the physical modeling is achieved by means of mass, energy and momentum balances, loss correlations and equations of state.
  • the undegraded model models an actual gas flow within the compressor by considering a stationary co-ordinate system and a rotating reference coordinate system for modeling thermodynamic transformations which occur within the gas flow in the compressor between inlet and outlet of gas path components such as inlet guide vanes, impeller, diffuser and return channel, for example.
  • a performance map provides an indication of the performance such as, for example, an achievable compression ratio and thermodynamic efficiencies as a function of an inlet gas flow and other manipulated variables such as rotational speed of the axle, inlet throttling and/or inlet pre-swirling, for example. It also may contain surge and choke lines that limit stable operative margins of the particular compressor.
  • Performance maps are either based on compressor rig test results or are predicted by a special computer program.
  • a performance map of a compressor can be scaled with respect to performance maps of comparable compressors or with respect to historical data of the compressor itself, respectively.
  • the compressor delivers compressed gas flow to a system such as, for example, an engine or plant to which it is connected by means of a pipeline system.
  • a working line is based on a locus of operating points that result from an interaction of the compressor and the system. Different working lines of a particular compressor can be observed when pressure drops inside the compressor are increased, for example by throttling of a valve placed on the pipeline system.
  • the present method is based on an offline calibration and validation of the undegraded model by using performance maps of, for example, a compressor model or a comparable compressor in an undegraded state.
  • calibration means adjusting the undegraded model with reference to performance maps and/or historical data based on a compressor in an undegraded state, such that the undegraded model is suitable for a simulation of data according to an undegraded state of the particular compressor.
  • the undegraded model can be calibrated by using data such as, for instance, geometry and process measurements of the particular compressor. If a calibrated model already exists, the model can be validated by using data of a performance map and/or historical data gathered by a pressure sensor of the particular compressor, for example.
  • a performance map of the particular compressor can be simulated i.e. calculated, wherein different running lines of the particular compressor may be simulated.
  • performance maps i.e. respective values of the performance maps of the undegraded model and the degradation-adaptive model
  • particular differences i.e. normalized deviations calculated by a difference of values derived by the undegraded model and values of the degradation-adaptive model divided by values of the undegraded model, for example, can be calculated for performance parameters such as pressure ratios and polytropic efficiencies, for instance.
  • Differences which have been calculated by comparing performance maps of the undegraded model and the degradation-adaptive model reveal potential degradations of a particular compressor and, therefore, information according to an actual efficiency of the particular compressor.
  • the comparison of performance maps derived by the undegraded model and the degradation-adaptive model, respectively, can be used to calculate an operating point which will lead to a decreased degradation and/or an optimized performance. Thereto, angles of inlet guide vanes or rotation speed of the axle of the particular compressor can be adjusted accordingly, for example.
  • the present method can also be used for a set of compressors, wherein each compressor is controlled by using particular models, with the objective of reducing overall energy costs for operation of the set of compressors .
  • the comparison may be used to obtain an optimal maintenance schedule for the particular compressor or set of compressors such that a performance of the compressor or set of compressors will not decrease under a distinct threshold.
  • the method can be used for an automatic identification and correction of failures of at least one sensor. By exploiting a physical and an analytical redundancy of a particular compressor and its respective model, particular failures can easily be detected and corrected, if necessary.
  • an optimization algorithm may be used which is based on a value calculated by a difference between input and output calculated by using real measurements on the one side and either by using the undegraded model or the degradation-adaptive model on the other side, wherein the undegraded model may be used for identification of ongoing failures in the running compressor and the degradation-adaptive model may be used for prediction of failures that arise in the future, for example .
  • sensors may be used to acquire these actual data.
  • Sensors that are suitable for such an acquisition may be a pressure sensor for measuring actual boost and/or gas flow within the compressor.
  • the present method may be used to prevent breakdowns or to increase efficiency of the combustion engine.
  • the present method may also be used for turbines or any other technical apparatus which is based on at least one compressor.
  • the present invention further relates to a diagnosis device for a compressor, the diagnosis device comprising a processor unit configured to compare characteris- tic values of performance maps of an undegraded model and a degradation-adaptive model of the compressor, wherein the undegraded model is set up in an undegraded state of the compressor and calibrated and validated offline by using historical data from a compressor running line of a comparable compressor or the compressor itself, and wherein the undegraded model is updated online with operative data of the respective compressor to obtain the degradation- adaptive model, and wherein the processor unit is configured to compare performance maps derived by simulations based on the degradation-adaptive model and the undegraded model, respectively, with each other in order to detect performance deviations of the respective compressor.
  • the present diagnosis device is suitable for detection of degradation and/or failures of a compressor in a plant or a combustion engine, for example.
  • diagnosis device is configured to be used to perform the method according to the present invention as described above.
  • the diagnosis device may further be incorporated in a control device of a plant, an engine or a respective vehicle, for example.
  • the diagnosis device may also be incorporated in a computer which may be connected to a combustion engine by a wire transfer protocol or by means of wireless communication .
  • a diagnosis of an engine or a respective compressor may be conducted via the Internet, such that data of a number of different compressors can be compared and analyzed in order to detect failures of the compressor and/or the engine.
  • Figure 1 shows a possible embodiment of a schematic overview of a framework for performance monitoring of a compressor by carrying out an embodiment of the method according to the present invention .
  • Figure 2 shows a schematic overview of a possible embodiment of an algorithm for updating a degradation-adaptive model according to the present invention.
  • Figure 3 shows a schematic overview of a possible embodiment of a framework for correction of data measured by sensors and calculated by a model as well as for estimation of optimal parameters for updating the degradation-adaptive model.
  • Figure 1 provides an overview of a framework which comprises a data filter 23 that filters data 22 from a compressor 21, such as a multistage intercooled air centrifugal compressor, based on a vector of input data u(t) such as for example rotating speed, and a vector of output data y(t) such as, for example, outlet temperature and pressure. Due to possible outliers, data 22 have to be filtered by filter 23, such that transformed, i.e. filtered data 22', namely u(k) and y(k) are derived. Then, the vector of input data u(k) is passed to an undegraded model 25, whereas output data y(k) and input data u(k) are both passed to a degradation-adaptive model 24.
  • a data filter 23 that filters data 22 from a compressor 21, such as a multistage intercooled air centrifugal compressor, based on a vector of input data u(t) such as for example rotating speed, and a vector of output data y(t) such as, for example
  • the undegraded model 25 is calibrated and validated offline using historical data from a performance map of the compressor 21 or a comparable compressor.
  • the undegraded model 25 comprises calculations of mass and energy balances as well as equations of state, ID flow models and mechanicals loss correlations for a gas flow in the compressor 21.
  • Input variables u(k) which may comprise measurements of inlet flow conditions and operative variables enter the undegraded model 25 as input.
  • the undegraded model 25 produces predictions for outlet flow characteristics in undegraded conditions which are then used for estimating performance parameters of the compressor 21 in undegraded conditions, namely a pressure ratio PR UD (k) and a polytropic efficiency n 0D (k) in an estimation step 28.
  • a threshold for maximum allowed degradation can be chosen optimally based on plant and/or engine set-ups e.g. discontinuity in trends shown in diagrams 30 may be representative of a maintenance action such as cleaning, for example.
  • Data 22 from gas path components of the compressor 21 are collected while compressor 21 is in operation and passed through filter 23 of the monitoring system 201.
  • the degradation-adaptive 24 model is based on the same data u(k) as the undegraded model 25 and updated, by filtered output data y(k) .
  • performance parameters are estimated based on output data y(k) resulting in pressure ratio PR(k) and polytropic efficiency (k), whereas an output of the undegraded model 25 which is based on input variables u(k) is used to calculate respective undegraded performance parameters H UD ( k) and PRuo(k).
  • Step 29 is based on the performance parameters estimated in step 28.
  • the performance parameters estimated in step 28 are used for calculation of diagrams 30.
  • a parameter trending and degradation detection algorithm is performed based on diagrams 30.
  • Diagrams 30 show, for example, actual process conditions that are monitored online. These process conditions may be determined by a ratio of a difference between a pressure ratio PR UD calculated based on outputs of the undegraded model 25 and a pressure ratio PR calculated based on filtered output data y(k) of the compressor 21, and the pressure ratio calculated based on outputs of the undegraded model 25 itself over time as shown in diagram 30_1. The same is applied to the polytropic efficiency ⁇ as shown in diagram 30_2. Thus, process conditions are shown with respect to output data y(t) of the compressor 21.
  • Performance deviations are monitored online, by use of diagrams 30, for assessing if maintenance actions are required. Estimated performance deviations are representative of a mechanical degradation because any other effect but operating time is taken into account by the undegraded model 25. Observation of trends of normalized deviation is a simple, reliable and sufficient mean for judging need of a maintenance intervention.
  • a gas flow through the compressor 21 may be used for calibration of the undegraded model 25 .
  • a gas flow through compressor 21 may be modeled by using the following equations (101, 102), wherein equation (102) defines empirical parts of a respective model based on general knowledge about compressors:
  • / is an algebraic system of equations representing physical parts of a respective model such as mass and energy balances, equations of state, a temperature profile, ID flow models and mechanical loss correlations
  • ⁇ is a vector of parameters, for example in main geometrical dimensions, loss parameters and fluid property parameters
  • q is an algebraic system of equations representing empirical parts of a respective model
  • ⁇ is a matrix of empirical parameters.
  • Variables z can not be measured directly, but they can be related to inputs through algebraic functions q that need to be empirically estimated for calibration of the undegraded model 25.
  • model parameters ⁇ ( . 33 which define a system response are recursively updated online.
  • a moving window for incorporating new data while keeping memory of previous model parameters is used.
  • the mentioned optimization problem considers a section of readings of duration T that is close to a cur- rent time instance t and, therefore, representative of a current status of degradation-adaptive model 24.
  • Parameters ⁇ , 33 are modified by an updating factor ⁇ which is bounded between an upper UB and a lower bound LB.
  • a duration of T determines how often the degrada- tion-adaptive model 24 is updated, while ⁇ ⁇ bounds define how much newly available data are trusted for updating degradation-adaptive model 24.
  • T, UB and LB are considered as tuning factors in the mentioned optimization problem. Tuning of T, UB and LB has to guarantee that the degradation- adaptive model 24 is able to adapt to slowly progressing degradation. The tuning does not have to be too fast also in order to prevent that an update of the degradation- adaptive model 24 is pursued with a non-representative section of readings that could for example contain readings from a temporary malfunctioning sensor.
  • the optimization algorithm shown in Figure 3 starts with an initialization step 41, wherein, for each steady state window, an optimization problem is initialized using actual measured values and parameters updated at a previous steady state window.
  • an optimizer 43 provides values of adaptive empirical parameters ⁇ 3 ⁇ 4; which may be used to model an effect of degradation, and a vector of input variables u k such as mass flow, for example.
  • Q k and u k may be used to obtain output variables y k by finding the roots of the physical nonlinear implicit algebraic model described previously.
  • the adaptive parameters Q k are measurable directly. In a first step, these parameters Q k have to be estimated during offline calibration of the undegraded model and then updated online to capture the slowly progressing mechanical degradation.
  • a weighted least square function (108) is evaluated in an evaluation step 44.
  • the weighted least square function (108) is based on a deviation of estimated values y and u of input and output variables and actual measurements 45 (y m and u m ) , for a window of time where respective readings are in steady state, wherein V is a covariance matrix for measured variables which are estimated using historical data.
  • V is a covariance matrix for measured variables which are estimated using historical data.
  • the optimizer 43 will select new values Q k+ i and u k+ i based on Q k and u k but limited to particular limits such that the following restrictions are met: Q min ⁇ Q ⁇ Q max and u min ⁇ i7 ⁇ u max .
  • the algorithm relies on predictions of the respective model for reconciling values of biased variables.
  • respective parameters have to be tightly constrained, to adapt only to slow changes in the particular compressor.
  • the degradation-adaptive model may be used to predict degradation and, therefore, performance of the compressor in the future. Based on the degradation- adaptive model, information can be extrapolated, i.e. simulated such that critical effects in the future may be predicted and corrected even before they arise by using, for example, a different running line of the compressor.

Abstract

The present invention relates to a method for online evaluation of operative range and performance of a compressor, the method comprising at least the following offline steps: setting up a digital undegraded model of the compressor in an undegraded state of the compressor; calibrating and validating the undegraded model using historical data from a compressor running line; calculating at least one undegraded performance map using the undgraded model; and at least the following online steps: calculating a degradation-adaptive model by updating the undegraded model with operative data of the compressor determined by at least one sensor of the compressor; calculating at least one actual performance map of an operative state of the compressor using the degradation-adaptive model; detecting malfunctions of the compressor by comparing the at least one undegraded performance map derived by the undegraded model and the at least one actual performance map derived by the degradation-adaptive model via at least one mathematical function. The present invention also refers to a corresponding diagnosis device.

Description

Method and device for online evaluation of a compressor
Technical field
[0001] The present invention refers to a method for online evaluation of a compressor.
[0002] The work leading to this invention has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement number 264940.
Description of prior art:
[0003] Compressors are used m a wide area such as for compressing fluids or gases to run, for example, a plant or an engine.
[0004] In case of using compressors for supercharging combustion engines an amount of air which is used for combustion within a cylinder of the engine is increased by using one or more compressors which are moved by a turbine that is in turn driven by either an electric engine, or by exhausting gases of the combustion engine itself.
[0005] A compressor can be degraded by the very high frequency of rotations of a turbine within the compressor and the very high temperatures of the exhausting gases which may lead to a temperature gradient within the compressor and, therefore, causing strains within the material of e.g. an impeller and/or the compressor itself. Further, parts of a compressor are exposed to a particular environment and, therefore, prone to slow changes such as fouling, erosion or corrosion. Thus, compressors are under high mechanical stress which may lead to a decreased performance of the compressor and/or a respective combustion engine or plant .
[0006] Traditional turbomachinery monitoring methods are based on system identification approaches that involve, for example, acoustic, visual or vibration monitoring in order to identify malfunctions of a respective compressor.
[0007] Another approach known in the state of the art is the use of model-based pattern recognition approaches which are trained on a set of training data and programmed to detect specific patterns which may be correlated to a malfunction .
[0008] Based on the state of the art, slow changes such as decreasing efficiency caused by, for example, fouling or erosion can not be efficiently detected.
[0009] Therefore, it is an object of the present invention to automatically identify performance problems of a compressor caused by degrading effects over long time periods, in order to prevent breakdowns and to modify maintenance schedules for improving efficiency of the compressor, respectively .
Summary of the invention
[0010] The above mentioned problem resolved by a method according to claim 1. [0011] According to the subject matter of claim 1, the method for online evaluation of an operative range and performance of a compressor comprises at least the following offline steps: a) setting up a digital undegraded model of the compressor in an undegraded state of the compressor; b) calibrating and validating the undegraded model offline using historical data from a compressor running line; c) calculating at least one undegraded performance map using the undegraded model; and at least the following online steps: d) calculating a degradation-adaptive model by updating the undegraded model with operative data of the compressor determined by at least one sensor of the compressor; e) calculating at least one actual performance map of an operative state of the compressor running line; f) detecting malfunctions of the compressor by comparing the at least one undegraded performance map derived by the undegraded model and the at least one actual performance map derived by the degradation- adaptive model via at least one mathematical function . [0012] For reasons of clarity, the terms "turbo charger" and "compressor" are summarized under the term "compressor" in the following.
[0013] The described method is based on an undegraded model and a degradation-adaptive model, wherein the undegraded model is set up in a first step offline. Thus, the undegraded model is set up in a non-operative state of a particular compressor, wherein the undegraded model is based on a physical modeling of the compressor. The physical modeling is achieved by means of mass, energy and momentum balances, loss correlations and equations of state. Based on these physical laws, the undegraded model models an actual gas flow within the compressor by considering a stationary co-ordinate system and a rotating reference coordinate system for modeling thermodynamic transformations which occur within the gas flow in the compressor between inlet and outlet of gas path components such as inlet guide vanes, impeller, diffuser and return channel, for example.
[0014] To set up the undegraded model, information regarding geometrical dimensions of radii and blade angles as well as a number of revolutions per minute of an axle, a mass flow rate and its respective suction conditions such as, for example, pressure, temperature and composition have to be available. Based on the physical modeling, intermediate and outlet intensive variables such as temperatures and pressures can be estimated. For calibration purposes, the prediction of outlet temperatures and pressures are compared with actual measurements from the compressor.
[0015] In order to create a degradation-adaptive model, process measurements of gas flow, pressure, temperatures and composition from the particular compressor are performed by respective sensors of the compressor and taken into account for converting the undegraded model into the degradation-adaptive model by updating it accordingly, as new measurements become available online i. e. in an operative state of the particular compressor. An update of the undegraded model by using measurements in order to obtain a degradation-adaptive model guarantees accuracy of the degradation-adaptive model and it permits an estimation of effects of mechanical degradation even away from a particular compressor running line, i.e. working line. Thus, by using the degradation-adaptive model, a performance of the particular compressor can be simulated, i.e. predicted for another running line, for example.
[0016] A performance map provides an indication of the performance such as, for example, an achievable compression ratio and thermodynamic efficiencies as a function of an inlet gas flow and other manipulated variables such as rotational speed of the axle, inlet throttling and/or inlet pre-swirling, for example. It also may contain surge and choke lines that limit stable operative margins of the particular compressor.
[0017] Performance maps are either based on compressor rig test results or are predicted by a special computer program. Alternatively, a performance map of a compressor can be scaled with respect to performance maps of comparable compressors or with respect to historical data of the compressor itself, respectively.
[0018] The compressor delivers compressed gas flow to a system such as, for example, an engine or plant to which it is connected by means of a pipeline system. A working line is based on a locus of operating points that result from an interaction of the compressor and the system. Different working lines of a particular compressor can be observed when pressure drops inside the compressor are increased, for example by throttling of a valve placed on the pipeline system.
[0019] The present method is based on an offline calibration and validation of the undegraded model by using performance maps of, for example, a compressor model or a comparable compressor in an undegraded state. Thus, calibration means adjusting the undegraded model with reference to performance maps and/or historical data based on a compressor in an undegraded state, such that the undegraded model is suitable for a simulation of data according to an undegraded state of the particular compressor. The undegraded model can be calibrated by using data such as, for instance, geometry and process measurements of the particular compressor. If a calibrated model already exists, the model can be validated by using data of a performance map and/or historical data gathered by a pressure sensor of the particular compressor, for example.
[0020] Based on the undegraded model which was set offline, comparisons can be made for detecting deviations of a performance level of the particular compressor.
[0021] Based on the degradation-adaptive model which comprises an actual state of the particular compressor, including also information about mechanical degradation caused by, for example, fouling or erosion, a performance map of the particular compressor can be simulated i.e. calculated, wherein different running lines of the particular compressor may be simulated. [0022] By comparing performance maps, i.e. respective values of the performance maps of the undegraded model and the degradation-adaptive model, particular differences, i.e. normalized deviations calculated by a difference of values derived by the undegraded model and values of the degradation-adaptive model divided by values of the undegraded model, for example, can be calculated for performance parameters such as pressure ratios and polytropic efficiencies, for instance. Differences which have been calculated by comparing performance maps of the undegraded model and the degradation-adaptive model reveal potential degradations of a particular compressor and, therefore, information according to an actual efficiency of the particular compressor.
[0023] The comparison of performance maps derived by the undegraded model and the degradation-adaptive model, respectively, can be used to calculate an operating point which will lead to a decreased degradation and/or an optimized performance. Thereto, angles of inlet guide vanes or rotation speed of the axle of the particular compressor can be adjusted accordingly, for example.
[0024] The present method can also be used for a set of compressors, wherein each compressor is controlled by using particular models, with the objective of reducing overall energy costs for operation of the set of compressors .
[0025] Further, the comparison may be used to obtain an optimal maintenance schedule for the particular compressor or set of compressors such that a performance of the compressor or set of compressors will not decrease under a distinct threshold. [0026] According to a further embodiment, the method can be used for an automatic identification and correction of failures of at least one sensor. By exploiting a physical and an analytical redundancy of a particular compressor and its respective model, particular failures can easily be detected and corrected, if necessary. Thereto, an optimization algorithm may be used which is based on a value calculated by a difference between input and output calculated by using real measurements on the one side and either by using the undegraded model or the degradation-adaptive model on the other side, wherein the undegraded model may be used for identification of ongoing failures in the running compressor and the degradation-adaptive model may be used for prediction of failures that arise in the future, for example .
[0027] Since the degradation-adaptive model becomes updated by actual data, sensors may be used to acquire these actual data. Sensors that are suitable for such an acquisition may be a pressure sensor for measuring actual boost and/or gas flow within the compressor.
[0028] In case of a compressor linked to combustion engine in a vehicle such as a turbocharger , for example, the present method may be used to prevent breakdowns or to increase efficiency of the combustion engine.
[0029] Further, the present method may also be used for turbines or any other technical apparatus which is based on at least one compressor.
[0030] The present invention further relates to a diagnosis device for a compressor, the diagnosis device comprising a processor unit configured to compare characteris- tic values of performance maps of an undegraded model and a degradation-adaptive model of the compressor, wherein the undegraded model is set up in an undegraded state of the compressor and calibrated and validated offline by using historical data from a compressor running line of a comparable compressor or the compressor itself, and wherein the undegraded model is updated online with operative data of the respective compressor to obtain the degradation- adaptive model, and wherein the processor unit is configured to compare performance maps derived by simulations based on the degradation-adaptive model and the undegraded model, respectively, with each other in order to detect performance deviations of the respective compressor.
[0031] The present diagnosis device is suitable for detection of degradation and/or failures of a compressor in a plant or a combustion engine, for example.
[0032] Further, the diagnosis device is configured to be used to perform the method according to the present invention as described above.
[0033] The diagnosis device may further be incorporated in a control device of a plant, an engine or a respective vehicle, for example.
[0034] The diagnosis device may also be incorporated in a computer which may be connected to a combustion engine by a wire transfer protocol or by means of wireless communication .
[0035] In case of wireless communication, a diagnosis of an engine or a respective compressor may be conducted via the Internet, such that data of a number of different compressors can be compared and analyzed in order to detect failures of the compressor and/or the engine.
Brief description of the drawings
Figure 1 shows a possible embodiment of a schematic overview of a framework for performance monitoring of a compressor by carrying out an embodiment of the method according to the present invention .
Figure 2 shows a schematic overview of a possible embodiment of an algorithm for updating a degradation-adaptive model according to the present invention.
Figure 3 shows a schematic overview of a possible embodiment of a framework for correction of data measured by sensors and calculated by a model as well as for estimation of optimal parameters for updating the degradation-adaptive model.
Detailed description
[0036] Figure 1 provides an overview of a framework which comprises a data filter 23 that filters data 22 from a compressor 21, such as a multistage intercooled air centrifugal compressor, based on a vector of input data u(t) such as for example rotating speed, and a vector of output data y(t) such as, for example, outlet temperature and pressure. Due to possible outliers, data 22 have to be filtered by filter 23, such that transformed, i.e. filtered data 22', namely u(k) and y(k) are derived. Then, the vector of input data u(k) is passed to an undegraded model 25, whereas output data y(k) and input data u(k) are both passed to a degradation-adaptive model 24. Further, the undegraded model 25 is calibrated and validated offline using historical data from a performance map of the compressor 21 or a comparable compressor. The undegraded model 25 comprises calculations of mass and energy balances as well as equations of state, ID flow models and mechanicals loss correlations for a gas flow in the compressor 21.
[0037] Input variables u(k) which may comprise measurements of inlet flow conditions and operative variables enter the undegraded model 25 as input. The undegraded model 25 produces predictions for outlet flow characteristics in undegraded conditions which are then used for estimating performance parameters of the compressor 21 in undegraded conditions, namely a pressure ratio PRUD(k) and a polytropic efficiency n0D(k) in an estimation step 28. By incorporating filtered output data y(k), degradation related estimation for pressure ratio PR(k) and polytropic efficiency (k) according to estimation step 28 is used for degradation detection and parameters trending in step 29. Based on parameter trending, as shown in diagrams 30, a threshold for maximum allowed degradation can be chosen optimally based on plant and/or engine set-ups e.g. discontinuity in trends shown in diagrams 30 may be representative of a maintenance action such as cleaning, for example.
[0038] Data 22 from gas path components of the compressor 21 are collected while compressor 21 is in operation and passed through filter 23 of the monitoring system 201. The degradation-adaptive 24 model is based on the same data u(k) as the undegraded model 25 and updated, by filtered output data y(k) . [0039] In the estimation step 28 performance parameters are estimated based on output data y(k) resulting in pressure ratio PR(k) and polytropic efficiency (k), whereas an output of the undegraded model 25 which is based on input variables u(k) is used to calculate respective undegraded performance parameters HUD ( k) and PRuo(k).
[0040] Step 29 is based on the performance parameters estimated in step 28. This means, the performance parameters estimated in step 28 are used for calculation of diagrams 30. In step 29 a parameter trending and degradation detection algorithm is performed based on diagrams 30. Diagrams 30 show, for example, actual process conditions that are monitored online. These process conditions may be determined by a ratio of a difference between a pressure ratio PRUD calculated based on outputs of the undegraded model 25 and a pressure ratio PR calculated based on filtered output data y(k) of the compressor 21, and the pressure ratio calculated based on outputs of the undegraded model 25 itself over time as shown in diagram 30_1. The same is applied to the polytropic efficiency η as shown in diagram 30_2. Thus, process conditions are shown with respect to output data y(t) of the compressor 21.
[0041] Performance deviations are monitored online, by use of diagrams 30, for assessing if maintenance actions are required. Estimated performance deviations are representative of a mechanical degradation because any other effect but operating time is taken into account by the undegraded model 25. Observation of trends of normalized deviation is a simple, reliable and sufficient mean for judging need of a maintenance intervention. [0042] For calibration of the undegraded model 25 a gas flow through the compressor 21 may be used. A gas flow through compressor 21 may be modeled by using the following equations (101, 102), wherein equation (102) defines empirical parts of a respective model based on general knowledge about compressors:
Figure imgf000014_0001
Figure imgf000014_0002
[0045] Therein: / is an algebraic system of equations representing physical parts of a respective model such as mass and energy balances, equations of state, a temperature profile, ID flow models and mechanical loss correlations, Θ is a vector of parameters, for example in main geometrical dimensions, loss parameters and fluid property parameters, ,. is a vector of intermediate variables z such as flow angle and pressure loss coefficient, q is an algebraic system of equations representing empirical parts of a respective model and Ω is a matrix of empirical parameters. Variables z can not be measured directly, but they can be related to inputs through algebraic functions q that need to be empirically estimated for calibration of the undegraded model 25.
[0046] One possible methodology for calibration of the undegraded model 25 to an existing performance map starts with an inverse calculation of intermediate variables for a number of selected calibration data-points due to equation 103.
[0047] yi-f{ui,0,zi) = Q→zi (103) [0048] Back-calculated values of intermediate variables z do not vary randomly with respect to a vector of in¬ puts values w, but follow defined patterns. A non-linear relation between ut and zf can be well described by empirical functions, wherein parameters of matrix Ω are tuned over a calibration data-set. A benefit of this approach is that if a performance map is missing, the undegraded model 25 can still be calibrated using data measurements as long as they are consistent with modeling assumptions. To guarantee consistence of data measurements with the modeling assumptions the respective data are transformed i.e. filtered, for example.
[0049] Independent from the online monitoring output data y(t), measurements of outlet flow conditions of compressor 21 are adopted for updating the degradation- adaptive model 24. Maps 27, for example performance maps showing a pressure ratio PR or a polytropic efficiency η with respect to a normalized volumetric flow in [%], are plotted in a complete operative range by using simulation and monitoring of performance maps in a simulation step 26 based on an output of the degradation-adaptive model 24 and the undegraded model 25. Thus, in contrast to the monitoring step 29, an effect of mechanical degradation can be es¬ timated also for other operating parameters away from an actual running line of the compressor 21. Thus, performance predictions of compressor 21 for a different running line become available.
[0050] Figure 2 shows an online adaptive modeling algorithm. Model parameters Ω, (i=l,...,s) 33 are updated recursively online as new operative variables Ui ( i=l , s ) 31 and output variables Υ ( i=l , s ) 32 become available. [0051] In order to match the degradation-adaptive model 24 to an actual degraded status, model parameters Ω(. 33 which define a system response are recursively updated online. Thus, a moving window for incorporating new data while keeping memory of previous model parameters is used. One possible mathematical formulation of a respective optimization problem is as follows:
[0052] Us = [u(k s_2)T)...u(kT+{s-2)T)) (104)
Figure imgf000016_0001
Figure imgf000016_0002
subject to LB<0s < UB ,
Ω,=Ω,_,(1 +^). (107)
[0053] The mentioned optimization problem considers a section of readings of duration T that is close to a cur- rent time instance t and, therefore, representative of a current status of degradation-adaptive model 24.
[0054] Parameters Ω, 33 are modified by an updating factor ι which is bounded between an upper UB and a lower bound LB. A duration of T determines how often the degrada- tion-adaptive model 24 is updated, while φί bounds define how much newly available data are trusted for updating degradation-adaptive model 24. T, UB and LB are considered as tuning factors in the mentioned optimization problem. Tuning of T, UB and LB has to guarantee that the degradation- adaptive model 24 is able to adapt to slowly progressing degradation. The tuning does not have to be too fast also in order to prevent that an update of the degradation- adaptive model 24 is pursued with a non-representative section of readings that could for example contain readings from a temporary malfunctioning sensor.
[0055] The optimization algorithm shown in Figure 3 starts with an initialization step 41, wherein, for each steady state window, an optimization problem is initialized using actual measured values and parameters updated at a previous steady state window. In an optimization step 42 an optimizer 43 provides values of adaptive empirical parameters Ω¾; which may be used to model an effect of degradation, and a vector of input variables uk such as mass flow, for example. Qk and uk may be used to obtain output variables yk by finding the roots of the physical nonlinear implicit algebraic model described previously. The adaptive parameters Qk are measurable directly. In a first step, these parameters Qk have to be estimated during offline calibration of the undegraded model and then updated online to capture the slowly progressing mechanical degradation.
[0056] Further, a weighted least square function (108) is evaluated in an evaluation step 44.
[0057] J = (y-y»>)V-l(y-ym)+
Figure imgf000017_0001
(108)
[0058] The weighted least square function (108) is based on a deviation of estimated values y and u of input and output variables and actual measurements 45 (ym and um) , for a window of time where respective readings are in steady state, wherein V is a covariance matrix for measured variables which are estimated using historical data. [0059] According to an optimization criterion for a Jacobean of objective function 46, which may be zero or below a threshold, the algorithm decides whether an optimal solution 47, which can be used to update the degradation- adaptive model was found or whether another iterative step based on steps 42 and 44 has to be carried out. If another iterative step has to performed, the optimizer 43 will select new values Qk+i and uk+i based on Q k and uk but limited to particular limits such that the following restrictions are met: Q min< Q < Q max and umin<i7 <umax. Based on the optimization algorithm, it is possible to correct data also when measurements are strongly biased. In situation where measurements are biased, the algorithm relies on predictions of the respective model for reconciling values of biased variables. However, respective parameters have to be tightly constrained, to adapt only to slow changes in the particular compressor.
[0060] Further, the degradation-adaptive model may be used to predict degradation and, therefore, performance of the compressor in the future. Based on the degradation- adaptive model, information can be extrapolated, i.e. simulated such that critical effects in the future may be predicted and corrected even before they arise by using, for example, a different running line of the compressor.

Claims

Claims
1. Method for online evaluation of operative range and performance of a compressor, the method comprising at least the following offline steps: setting up a digital undegraded model of the compressor in an undegraded state of the compressor; calibrating and validating the undegraded model using historical data from a compressor running line; calculating at least one undegraded performance map using the undgraded model; and at least the following online steps: determining operative data of the compressor by at least one sensor of the compressor; calculating a degradation-adaptive model by updating the undegraded model with the operative data of the compressor determined by the at least one sensor of the compressor; calculating at least one actual performance map of an operative state of the compressor using the degradation- adaptive model; detecting malfunctions of the compressor by comparing the at least one undegraded performance map derived by the undegraded model and the at least one actual performance map derived by the degradation-adaptive model via at least one mathematical function.
2. Method according to claim 1, wherein an optimization algorithm is performed that is based on a physical and analytical redundancy of the compressor for an automatic identification and correction of faults of the at least one sensor .
3. Method according to claim 1 or 2, wherein the compressor is attached to a combustion engine.
4. Method according to claim 3, wherein the compressor is chosen as a turbo charger.
5. Method according to any of the preceding claims, wherein the undegraded model is calculated based on at least the following mathematical assumptions: mass and energy balances, equations of state, ID flow models and mechanical loss correlations .
6. Method according to any of the preceding claims, wherein a pressure-sensor for measuring an actual airflow is chosen as the at least one sensor.
7. Method according to any of the preceding claims, wherein the outcome of the comparison of the undegraded model and the degradation-adaptive model is used for diagnosis of an engine failure.
8. Diagnosis device for a compressor comprising a processor unit configured to compare characteristic values of performance maps of an undegraded model and a degradation- adaptive model of the compressor, wherein the undegraded model is set up in an undegraded state of the compressor and calibrated and validated offline by using historical data from a compressor running line and wherein the undegraded model is updated online with operative data of the compressor to obtain a degradation-adaptive model and wherein the processor unit is configured to compare performance maps derived by simulations based on the degradation- adaptive model to performance maps derived by the undegraded model in order to detect performance deviations of the compressor.
9. Diagnosis device configured to be used for carrying out the method according to any one of claims 1 to 7.
PCT/EP2015/000675 2014-03-31 2015-03-30 Method and device for online evaluation of a compressor WO2015149928A2 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
EP14001184.2 2014-03-31
EP14001184 2014-03-31

Publications (2)

Publication Number Publication Date
WO2015149928A2 true WO2015149928A2 (en) 2015-10-08
WO2015149928A3 WO2015149928A3 (en) 2015-12-30

Family

ID=50472965

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2015/000675 WO2015149928A2 (en) 2014-03-31 2015-03-30 Method and device for online evaluation of a compressor

Country Status (1)

Country Link
WO (1) WO2015149928A2 (en)

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3255512A3 (en) * 2016-06-07 2018-01-03 General Electric Company Adaptive model-based method to quantify degradation of a power generation system
WO2018009546A1 (en) * 2016-07-07 2018-01-11 Aspen Technology, Inc. Computer system and method for the dynamic construction and online deployment of an operation-centric first-principles process model for predictive analytics
EP3599523A1 (en) * 2018-07-24 2020-01-29 Siemens Aktiengesellschaft Detection of a calibration condition of at least one unit
GB2578332A (en) * 2018-10-24 2020-05-06 Centrica Plc Method of detecting an operating state of an appliance
EP3764183A1 (en) * 2019-07-12 2021-01-13 Siemens Aktiengesellschaft Managing health condition of a rotating system
US11630446B2 (en) 2021-02-16 2023-04-18 Aspentech Corporation Reluctant first principles models
US11754998B2 (en) 2019-10-18 2023-09-12 Aspentech Corporation System and methods for automated model development from plant historical data for advanced process control
US11782401B2 (en) 2019-08-02 2023-10-10 Aspentech Corporation Apparatus and methods to build deep learning controller using non-invasive closed loop exploration
US11853032B2 (en) 2019-05-09 2023-12-26 Aspentech Corporation Combining machine learning with domain knowledge and first principles for modeling in the process industries
US11934159B2 (en) 2018-10-30 2024-03-19 Aspentech Corporation Apparatus and methods for non-invasive closed loop step testing with controllable optimization relaxation

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6823675B2 (en) * 2002-11-13 2004-11-30 General Electric Company Adaptive model-based control systems and methods for controlling a gas turbine
US7505844B2 (en) * 2005-11-18 2009-03-17 General Electric Company Model-based iterative estimation of gas turbine engine component qualities
US20130024179A1 (en) * 2011-07-22 2013-01-24 General Electric Company Model-based approach for personalized equipment degradation forecasting

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
None

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9896960B2 (en) 2016-06-07 2018-02-20 General Electric Company Adaptive model-based method to quantify degradation of a power generation system
EP3255512A3 (en) * 2016-06-07 2018-01-03 General Electric Company Adaptive model-based method to quantify degradation of a power generation system
WO2018009546A1 (en) * 2016-07-07 2018-01-11 Aspen Technology, Inc. Computer system and method for the dynamic construction and online deployment of an operation-centric first-principles process model for predictive analytics
US10990067B2 (en) 2016-07-07 2021-04-27 Aspen Technology, Inc. Computer system and method for the dynamic construction and online deployment of an operation-centric first-principles process model for predictive analytics
EP3599523A1 (en) * 2018-07-24 2020-01-29 Siemens Aktiengesellschaft Detection of a calibration condition of at least one unit
GB2578332A (en) * 2018-10-24 2020-05-06 Centrica Plc Method of detecting an operating state of an appliance
GB2578332B (en) * 2018-10-24 2021-07-28 Centrica Plc Method of detecting an operating state of an appliance
US11934159B2 (en) 2018-10-30 2024-03-19 Aspentech Corporation Apparatus and methods for non-invasive closed loop step testing with controllable optimization relaxation
US11853032B2 (en) 2019-05-09 2023-12-26 Aspentech Corporation Combining machine learning with domain knowledge and first principles for modeling in the process industries
WO2021009042A1 (en) 2019-07-12 2021-01-21 Siemens Aktiengesellschaft Managing health condition of a rotating system
US11933838B2 (en) 2019-07-12 2024-03-19 Siemens Aktiengesellschaft Managing health condition of a rotating system
EP3764183A1 (en) * 2019-07-12 2021-01-13 Siemens Aktiengesellschaft Managing health condition of a rotating system
US11782401B2 (en) 2019-08-02 2023-10-10 Aspentech Corporation Apparatus and methods to build deep learning controller using non-invasive closed loop exploration
US11754998B2 (en) 2019-10-18 2023-09-12 Aspentech Corporation System and methods for automated model development from plant historical data for advanced process control
US11630446B2 (en) 2021-02-16 2023-04-18 Aspentech Corporation Reluctant first principles models

Also Published As

Publication number Publication date
WO2015149928A3 (en) 2015-12-30

Similar Documents

Publication Publication Date Title
WO2015149928A2 (en) Method and device for online evaluation of a compressor
Tsoutsanis et al. A component map tuning method for performance prediction and diagnostics of gas turbine compressors
EP2390471B1 (en) Blade monitoring system and method
CN110717219B (en) Method and device for acquiring inlet flow of air compressor in complete machine state of aero-engine
JP2009162230A (en) Method and system for materializing real-time comparison with using alternative control system on turbine
Cortinovis et al. Online performance tracking and load sharing optimization for parallel operation of gas compressors
JP2010144727A (en) System and method for monitoring rotor blade health
JP2018204604A (en) Systems and methods for icing detection of compressors
JP2021139362A (en) Adjustment of operating limit (ol) threshold for compressor of gas turbine system based on mass flow loss
US20160365735A1 (en) Systems and Methods for Power Plant Data Reconciliation
CN107203364B (en) Prediction and identification method for full-working-condition characteristics of gas compressor
US11460818B2 (en) Evaluation apparatus, evaluation system, and evaluation method
RU2727839C2 (en) Method and system of machine control
Zaccaria et al. A model-based solution for gas turbine diagnostics: Simulations and experimental verification
MX2015005729A (en) A method for operating a compressor in case of failure of one or more measure signal.
CN102619774A (en) Surge control method of centrifugal compression equipment
JP2019522143A (en) Adaptive surge prevention control system and method
Cicciotti et al. Online performance monitoring of industrial compressors using meanline modelling
Liu et al. Nonlinear model based diagnostic of gas turbine faults: a case study
US20150276548A1 (en) Condition monitoring and analytics for machines
CN111275320B (en) Performance adjustment data processing method, system and storage medium of generator set
Trawiński Development of flow and efficiency characteristics of an axial compressor with an analytical method including cooling air extraction and variable inlet guide vane angle
JP2022517504A (en) Gas networks and methods for detecting failures in gas networks under pressure or vacuum
CN113677972A (en) Gas network and method for simultaneous detection of leaks and blockages in a gas network under pressure or vacuum
Indarti et al. Integration of Technology Capability for Performance Diagnostics of MS7001EA Using PYTHIA

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 15722045

Country of ref document: EP

Kind code of ref document: A2