EP4097688A1 - Sensorsystem und verfahren zum erkennen eines zustands zumindest einer maschine - Google Patents
Sensorsystem und verfahren zum erkennen eines zustands zumindest einer maschineInfo
- Publication number
- EP4097688A1 EP4097688A1 EP21700767.3A EP21700767A EP4097688A1 EP 4097688 A1 EP4097688 A1 EP 4097688A1 EP 21700767 A EP21700767 A EP 21700767A EP 4097688 A1 EP4097688 A1 EP 4097688A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- machine
- data sets
- evaluation unit
- measured values
- sensor system
- Prior art date
- Legal status (The legal status 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 status listed.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M15/00—Testing of engines
- G01M15/14—Testing gas-turbine engines or jet-propulsion engines
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
- G07C3/00—Registering or indicating the condition or the working of machines or other apparatus, other than vehicles
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F02—COMBUSTION ENGINES; HOT-GAS OR COMBUSTION-PRODUCT ENGINE PLANTS
- F02C—GAS-TURBINE PLANTS; AIR INTAKES FOR JET-PROPULSION PLANTS; CONTROLLING FUEL SUPPLY IN AIR-BREATHING JET-PROPULSION PLANTS
- F02C7/00—Features, components parts, details or accessories, not provided for in, or of interest apart form groups F02C1/00 - F02C6/00; Air intakes for jet-propulsion plants
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/09—Supervised learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/091—Active learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
Definitions
- the present disclosure relates in particular to a sensor system and a method for recognizing a state of at least one machine, as well as to a computer program product.
- analysis tools can be provided which allow automatic status detection for large amounts of data. Such analysis tools typically have to be adapted to a specific purpose, which in many cases is too time-consuming.
- Kl artificial intelligence
- the object of the present invention is to enable an improved detection of the state of machines.
- a sensor system for recognizing a state of at least one machine.
- the sensor system comprises one or more sensors for acquiring measured values of at least one machine, at least one communication interface and an evaluation unit.
- the evaluation unit is set up to acquire a plurality of data sets, each of which includes measured values from the one or more sensors. Furthermore, the evaluation unit is set up to select part of the data records by means of active learning and to provide the selected part of the data records to the at least one communication interface.
- Active learning is a special case of machine learning in which a (learning) algorithm is designed to query a user (or another information source) (interactively) in order to obtain desired results at new data points.
- a (learning) algorithm is designed to query a user (or another information source) (interactively) in order to obtain desired results at new data points.
- particularly relevant and / or decision-relevant data records can be used specifically for recognizing the state of the at least one machine, which enables an improved detection of the state of machines.
- a measure of the relevance of one or more data records is determined. This enables, for example, an optimized selection of training data sets for a machine learning model, which can then automatically and precisely monitor the condition of at least one machine after a particularly short training period.
- the sensors are each designed to measure a physical variable, for example to measure a temperature, a speed, a rotational speed or a pressure.
- the measured values can be recorded in the form of time series data.
- the evaluation unit receives several measured values that follow one another in time from one sensor or several sensors.
- the evaluation unit includes, for example, one or more computers.
- the at least one communication interface can comprise a software interface and / or a hardware interface.
- the data records that are provided at the at least one communication interface and selected by means of active learning allow the state of the at least one machine to be recognized, for example by a user.
- the evaluation unit can be set up to specify a sequence for the selected data records by means of active learning. In this way, the most relevant data records can be processed first, so that a particularly rapid improvement in the status recognition can be achieved.
- the evaluation unit can further be set up to subdivide the multiple data records with respect to the at least one parameter of the data records into at least two groups by means of a separation line, and to carry out the selection of the part of the data records based on the distance of the parameters of the respective data records from the separation line .
- the evaluation unit is set up to receive a classification of the data records provided at the at least one communication interface with regard to the state of the at least one machine.
- the classification includes two or three choices.
- a choice is made from two answers (e.g. A or B) or from three answers (e.g. A, B or C).
- the classification corresponds, for example, to the result of a yes / no decision or a decision between the options A (eg yes), B (eg no) and C (eg “unknown” or “undefined”).
- the evaluation unit receives the classification in each case in the form of classification data units.
- the classification data units each include, for example, the indication “yes” or “no” or some other indication of positive or negative, for example 1 or 0.
- the evaluation unit can be set up to analyze the classified data sets, in particular by means of machine learning, in order to record at least one parameter of the respective data set (eg a minimum and / or a maximum and / or a standard deviation).
- the evaluation unit is set up to use the recorded at least one parameter of the classified data sets in order to train a machine learning model. Since the training of the machine learning model is based on the selection of the part of the data sets made by means of active learning, a significantly improved quality of the training and consequently the decision-making using the machine learning model trained in this way is possible. Such training also allows extensive automation. Furthermore, the amount of data required for training the machine learning models can be reduced, since optimal training data sets can be selected. Furthermore, a simple adaptation to a wide variety of different applications is possible. An adaptation of the machine learning model to different use cases beyond the training is not necessary for many use cases.
- the machine learning model For training the machine learning model, for example, properties of the respective selected part of the data are extracted in the form of the at least one parameter (e.g. a maximum value, a minimum value, a median, a mean value, a variance or the like), the training being carried out based on these parameters. At least one of the parameters can be a statistical parameter.
- the machine learning model can be or comprise a classification model. In particular, the machine learning model can be or comprise an artificial neural network.
- the evaluation unit is set up to use the trained machine learning model to classify further data sets with measured values, in particular to recognize the state of at least one machine.
- a particularly reliable detection of the state of a machine can be carried out automatically with an optimally trained machine learning model.
- the at least one machine is, for example, a gas turbine engine, in particular a multiplicity of gas turbine engines.
- gas turbine engines in particular, it is often desirable to detect a deteriorating state of a sensor or a component monitored by a sensor as early as possible, which is made possible by the machine learning models trained as described above.
- a method for recognizing a state of at least one machine comprises generating measured values of the at least one machine by means of one or more sensors; the acquisition, with an evaluation unit, of a plurality of data sets with measured values from the one or more sensors; selecting, with the evaluation unit and by means of active learning, a part of the data sets; and providing the selected part of the data records on at least one communication interface for recognizing the state of the at least one machine.
- the sensor system can be used according to any configuration described herein.
- the trained machine learning model can be used to classify further (in particular not yet classified) data sets with measured values.
- the corresponding machine can be serviced depending on the classification of the further data records.
- the evaluation unit can generate a corresponding command. Alternatively or additionally, a message is sent.
- a computer program product comprising instructions which, when executed by one or more processors, cause the one or more processors to perform the steps of the method according to any embodiment described herein to execute.
- FIG. 1 shows an aircraft in the form of an aircraft with several gas turbine engines
- FIG. 2 is a side sectional view of a gas turbine engine
- FIG. 3 shows a sensor system for recognizing a state of at least one machine
- FIG. 4 details of the system according to FIG. 3;
- FIGS. 5 to 8 different examples of measured values
- FIG. 9 shows a method for training a machine learning model
- FIG. 10 shows an embodiment of an evaluation unit of the system according to FIG. 3;
- FIGS 11A-11C details of active learning.
- FIG. 1 shows an aircraft 8 in the form of an aircraft.
- the aircraft 8 comprises a plurality of gas turbine engines 10.
- FIG. 2 shows one of the gas turbine engines 10 of the aircraft 8 with a main axis of rotation 9.
- the gas turbine engine 10 comprises an air inlet 12 and a fan 23 which generates two air flows: a core air flow A and a bypass air flow B.
- the gas turbine engine 10 comprises a core 11 which the core air flow A.
- the core engine 11 comprises, in axial flow order, a low-pressure compressor 14, a high-pressure compressor 15, a combustion device 16, a high-pressure turbine 17, a low-pressure turbine 19 and a core thrust nozzle 20.
- An engine nacelle 21 surrounds the gas turbine engine 10 and defines a bypass duct 22 and a bypass thrust nozzle 18.
- the bypass air flow B flows through the bypass duct 22.
- the fan 23 is attached to the low-pressure turbine 19 via a shaft 26 and an epicyclic planetary gear 30 and is driven by it.
- the core air flow A is accelerated and compressed by the low-pressure compressor 14 and passed into the high-pressure compressor 15, where further compression takes place.
- the compressed air expelled from the high pressure compressor 15 is directed into the combustion device 16, where it is mixed with fuel and the mixture is burned. The resulting ones are called
- Combustion products then propagate through the high pressure and low pressure turbines 17, 19 and thereby drive them before they are ejected through the nozzle 20 to provide a certain thrust.
- the high pressure turbine 17 drives the high pressure compressor 15 through a suitable connecting shaft 27.
- the fan 23 generally provides the majority of the thrust.
- the epicyclic planetary gear 30 is a reduction gear.
- Low-pressure compressors can be understood to include the turbine stage with the lowest pressure or the compressor stage with the lowest pressure (ie that they do not include the fan 23) and / or the turbine and compressor stage which are interconnected by the connecting shaft 26 at the lowest speed in the engine (ie that it does not include the transmission output shaft which drives the fan 23), mean.
- the “low pressure turbine” and “low pressure compressor” referred to here may alternatively be known as the “medium pressure turbine” and “medium pressure compressor”.
- the fan 23 may be referred to as a first compression stage or compression stage with the lowest pressure.
- gas turbine engines to which the present disclosure may find application may have alternative configurations.
- such engines can have an alternative number of compressors and / or turbines and / or an alternative number of connecting shafts.
- the gas turbine engine shown in Figure 2 has a split flow nozzle 20, 22, which means that the flow through the bypass duct 22 has its own nozzle which is separate from the engine core nozzle 20 and radially outward therefrom.
- this is not limiting and any aspect of the present disclosure may also apply to engines in which the flow through the bypass duct 22 and the flow through the core 11 are in front of (or upstream) a single nozzle, which may be referred to as a mixed flow nozzle, mixed or combined.
- One or both nozzles can have a fixed or variable range.
- gas turbine engine such as a gas turbine engine.
- B in an open rotor (in which the fan stage is not surrounded by an engine nacelle) or a turboprop engine.
- the geometry of the gas turbine engine 10 and components thereof is or are defined by a conventional axis system which has an axial direction (which is aligned with the axis of rotation 9), a radial direction (in the direction from bottom to top in Figure 2) and a circumferential direction (perpendicular to the view in Figure 2) includes.
- the axial, radial and circumferential directions are perpendicular to one another.
- sensors are arranged on the gas turbine engine 10, of which several sensors 60-62 arranged at different points on the gas turbine engine 10, specifically temperature sensors for measuring temperatures, are shown here by way of example.
- FIG. 3 shows a sensor system 50.
- the sensor system 50 comprises the sensors 60-62 arranged in the present case on the gas turbine engine 10, an evaluation unit 52 and at least one communication interface 58, 59, in the present case two
- the evaluation unit 52 comprises, for example, one or more (adjacent or spaced apart) computers.
- the evaluation unit 52 is set up to receive several data sets, each with measured values from the sensors 60-62, and a part by means of active learning of the data records to be selected. Furthermore, the evaluation unit 52 is set up to provide the selected part of the data records for recognizing the state of at least one machine, in this case the gas turbine engine 10, to the communication interface 58.
- a memory 53 of the evaluation unit 52 there is at least one
- Machine learning model 51 is stored, in particular a plurality of machine learning models 51 are stored or can be stored. The several
- Machine learning models 51 can represent multiple instances of the same machine learning model.
- the evaluation unit 52 is a
- Communication interface 58 communicatively coupled to sensors 60-62 in order to acquire data, specifically measured values thereof. Furthermore, the evaluation unit 52 is communicatively coupled to at least one interface 81 (via the communication interface 58 or a further communication interface). In the present example, the interface 81 is graphical
- GUI User interfaces
- a user can classify one or more data record (s) selected by means of active learning.
- the user can also select the part of the measured values of the respective data record that is the reason for the selected classification. For example, the user can decide whether a respective data record includes measured values that indicate a specific condition of the gas turbine engine 10, such as the wear or defect of a component.
- the sensor system 50 can receive this classification (and optionally the selected part of the measured values) of the respective data record via the communication interface 58 and thus train the machine learning model 51.
- the machine learning models 51 are designed for machine learning and in the present example include a random forest and / or an artificial neural network.
- the sensor system 50 also includes a further machine learning model 57, which will be explained in more detail below. Based on the trained machine learning models 51, 57, data sets with Measured values are classified in order to make automated data-driven decisions, e.g. to trigger maintenance work.
- Instructions 54 which are part of a computer program product which cause a processor 55 of evaluation unit 52 to carry out the method shown in FIG. 9 (or at least part of it, in particular at least steps S11 to S13) are also stored in memory 53.
- the memory 53 is, for example, a non-volatile memory.
- the processor 55 includes, for example, a CPU, a GPU, and / or a tensor processor.
- the evaluation unit 52 further comprises a selector 56 (e.g. stored in the memory 53).
- the selector 56 is designed for active learning.
- the selector 56 is set up, for example, to select one or more data records (as a subset) from a large number of data records based on a predetermined rule, e.g. to select the data record that is to be used for the next classification, e.g. for which the selector 56 is the largest Probability is determined that this data set has the greatest training effect for the machine learning model 51.
- the selector 56 provides this data record to the interface 81 via the communication interface 58.
- the selector 56 can provide selected data sets via a software-based communication interface 59, e.g. to one or more machine learning models 51, 57.
- the selector 56 determines, for example, an order for the selected data sets.
- the evaluation unit 52 is optionally stationed on the ground and the gas turbine engine 10 can be moved relative thereto.
- FIG. 4 shows further details of the sensor system 50.
- measured values from the sensors 60-62 are stored in the form of a large number of time series and as raw data.
- the time series originate, for example, from several flights of the gas turbine engine 10, from the several gas turbine engines 10 of the aircraft 8 and / or from gas turbine engines 10 of several aircraft 8 (or, in general, from several machines).
- the transmission from the sensors 60-62 to the database 100 takes place, for example, via a data cable or wirelessly, for example via GSM or another mobile communication standard, in particular via the communication interface 58.
- the database 100 is stored in the memory 53, for example.
- the data stored in the database 100 are processed and stored in a further database 101, which can also be a transient data flow.
- a further database 101 can also be a transient data flow.
- uninteresting data cannot be transferred in order to simplify further processing.
- the measured values are further processed and stored in a further database 102 in order to carry out an analysis of the measured values with regard to suitable time series.
- This analysis takes place in block 117.
- data sets with measured values are selected from a larger number of data sets.
- suitable candidates are selected from data records, in particular each with a time series from a sensor 60-62 or with in each case a plurality of time series (in particular spanning the same time period) from a plurality of the sensors 60-62.
- the selection of the part of the data records is done by means of active learning.
- the selected part of the data records is provided to the communication interface 58 for recognizing the state of the at least one machine 10.
- measured values in certain time segments can allow particularly precise conclusions to be drawn about the state of the sensor or a machine monitored by the sensor.
- certain signatures in the data can represent an indication of a deteriorating state of the sensor or of a component that is monitored or can be monitored with it.
- the selected candidates or pointers to them are optionally stored in a database 110.
- an import script retrieves these candidates from the database 102 (or the database 101) in block 118 and provides them to a block 111 (optionally via a further database 106).
- a classification data unit and a selected part of the measured values of the respective candidate are recorded for all or for some of the candidates.
- the classification data units indicate a classification of the candidate into one of several predetermined classes.
- the classification data units and / or the selected parts of the measured values are provided by one or more users. This takes place, for example, via the interface 81 and / or the communication interface 58.
- the classification data units and selected parts of the candidates are stored in a database 108 and provided to a block 112.
- an instance of the machine learning model 51 is trained (e.g. for each user) based on the classification data units and selected parts of the candidates that were provided by the user.
- properties of the selected part of the measured values are extracted in the form of parameters.
- the extracted parameters and / or values calculated from them are the input parameters for the training. Examples of such parameters will be explained further below in connection with FIG.
- the data stored in the database 108 are provided to a block 113, which can also access the database 107.
- the (optional) superordinate machine learning model 57 is generated in block 113.
- the superordinate machine learning model 57 optionally corresponds to that
- Machine learning model 51 is e.g. with the (optionally weighted and / or selected) input parameters of several instances of the
- Machine learning model 51 trained.
- the superordinate machine learning model 57 and / or its input parameters is / are stored in a database 109 (which is, for example, stored in memory 53 is deposited).
- Machine learning model 57 shown on a user interface.
- the database 103 comprises the data of the database 102 to which optional selection or correction scripts have been applied.
- the databases 102 and 103 instead of the databases 102 and 103, only the database 102 is provided.
- the higher-level machine learning model 57 is applied to the measured values in the database 103 (or 102) in order to classify the measured values.
- the results of the classification from block 115 are stored in a database 104, optionally also data from the database 103 (or 102).
- data-driven decisions are made, e.g. the execution of maintenance work is triggered.
- the execution of maintenance work is triggered.
- a message is generated and transmitted, e.g. by email, indicating a decision.
- the data on which the decisions are based are optionally stored in a database 105.
- the databases 100 to 104 (which can also be logical steps through a data flow) are optionally part of an engine equipment health management, EHM, of the gas turbine engine 10 and / or are stored in the memory 53.
- the database 105 can be stationed on the ground, for example.
- the databases 100, 101, 102, 103, 104 and / or 105 can have separate physical memories or, alternatively, can be databases of a logical architecture, with, for example, several or all of the databases having the same physical memory .
- One or more of the blocks 111 to 118, in particular all of the blocks 111 to 118 can be stored in the form of instructions 54 in the memory 53 and can be executed by the processor 55.
- FIG. 5 shows exemplary measured values 70 in the form of time series data. A large number of measured values are plotted against time. Specifically, the measured values indicate a (first) temperature difference, which can be determined using two spaced apart sensors of a machine, in this case a diesel engine (alternatively e.g. analogously from one or two of the sensors 60-62) in the form of temperature sensors and has been determined here.
- a (first) temperature difference can be determined using two spaced apart sensors of a machine, in this case a diesel engine (alternatively e.g. analogously from one or two of the sensors 60-62) in the form of temperature sensors and has been determined here.
- FIG. 6 shows, by way of example, further measured values 70 in the form of time series data.
- a large number of measured values are plotted against time over the same period of time as the measured values in FIG. 5.
- the measured values in FIG Diesel engine can be determined in the form of temperature sensors and has been determined here, namely a different pair of sensors 60-62 than in FIG. 5.
- a selected part 71 of the measured values 70 is also illustrated in FIG. 6.
- the selected part 71 comprises conspicuous areas of the measured values. When measured values are conspicuous depends on the respective application. In the present example, values of the temperature difference below a certain limit and strong fluctuations in the values are noticeable.
- the selected part 71 generally comprises one or more temporal subsections of the measured values 70 (along the X axis). Optionally, the selected part 71 also includes a restriction along the Y-axis.
- FIG. 7 illustrates exemplary parameters that can be calculated from an exemplary selected part 71 of measured values 70.
- the parameters can be, for example, a maximum value, a minimum value, a median, a mean value, a variance, the sum of the squared individual values, the length of the selected part in time direction, a Autocorrelation or a parameter derived from it, the number of values above or below the mean value, the longest time interval above or below the mean value, the sum of the changes in the slope sign, a slope, a standard deviation and / or a number of peaks .
- Some of these parameters are highlighted graphically in FIG.
- One or more, for example all of these parameters can be used as input parameters for training the corresponding machine learning model 51.
- ratios of the parameters mentioned can be formed and used as input parameters for the training, for example mean value / variance, length / sum of the squared individual values or other ratios.
- FIG. 8 illustrates that time series data from several sensors can optionally be plotted in a multidimensional manner (here two-dimensionally), so that the selected part 71 of the measured values 70 can be selected in a multidimensional manner.
- a multidimensional manner here two-dimensionally
- several measured values that are correlated with one another can show particularly clear abnormalities, which can then be selected particularly easily and precisely.
- a point in the multidimensional representation corresponds to several different measured values at the same point in time.
- clusters of data points are determined in the multi-dimensional representation and e.g. their distances from one another and / or sizes, e.g. radii and / or their number of data points contained therein.
- FIG. 9 shows a method for classifying measured values, comprising the following steps:
- Step S1 Providing a trained machine learning model, in particular a trained superordinate machine learning model 57.
- a method for training the machine learning models 51 comprising steps S10 to S14:
- Step S10 generating measured values of at least one machine, in particular at least one gas turbine engine, by means of the one or more sensors 60-62, the measured values 70 being recorded in particular in the form of time series data and in particular displaying measured values from one or more gas turbines 10.
- Step S11 acquisition, by the evaluation unit 52, of data sets with measured values 70 obtained by means of the one or more sensors 60-62.
- Step S12 Select, with the evaluation unit 52 and by means of active learning, a part of the data records.
- Step S13 Providing, in particular by means of the evaluation unit 52, the selected part of the data records on the communication interface 58 (in particular for display on the interface 81) for recognizing the state of the at least one machine.
- Step S13 can further include: receiving, by evaluation unit 52, of classification data units relating to the measured values 70, the classification data units received by evaluation unit 52 relating to the data records provided at communication interface 58. Furthermore, step S13 comprises receiving, by the evaluation unit 52 and for each of the classification data units, a selected part 71 of the measured values 70.
- Step S14 Training, by means of the evaluation unit 52, of one or more machine learning models 51 based on the classification data units and the selected parts 71 of the measured values 70, the machine learning models 51 comprising, for example, an artificial neural network.
- machine learning models 51 e.g. several instances of the same type of machine learning model 51 are optionally trained (e.g. in that the above steps are each carried out by several users) and a superordinate machine learning model 57 is created from the several
- Step S2 comprises the classification, by the evaluation unit 52, of data records with measured values 70 acquired by means of the one or more sensors 60-62 using the at least one machine learning model 51 and / or the superordinate machine learning model 57.
- the optional step S3 comprises the generation, by the evaluation unit 52 and based on the classification of the data records of the measured values 70, of a command which indicates that maintenance work is being carried out.
- FIG. 10 shows schematically an optional embodiment of the evaluation unit 52.
- the database 102 is provided in which the data sets with the measured values obtained from the sensors 60-62 are stored.
- the data records can come from IoT devices (Internet of Things), from an engine control unit ECU or any other device for collecting physical data.
- each data record has measured values that each span an equally long period of time.
- Each data set can comprise measured values from one or, in particular, several sensors 60-62.
- the data records stored in the database 102 are provided to the block 117 via a communication link K1 (which is, for example, the block 117 explained in connection with FIG. 4).
- the block 117 provides (in particular successively) a part of the data records stored in the database 102 as candidates to the interface 81 via a communication connection K2.
- the measured values of the respective candidates are displayed via the interface 81 (communication connection K3) so that a user can classify them by making appropriate entries via the interface 81.
- the user selects a part of the measured values of the respective data record which comprises fewer measured values than the entire corresponding data record. This is the part of the measured values that gives rise to the respective classification.
- the user can be an expert who, based on their experience with a particularly good hit rate, identifies anomalies in the Can recognize records (but is unable to classify the potentially enormous amounts of records from database 102).
- An anomaly can be, for example, when a threshold value is exceeded, a specific trend, a sudden change or the like.
- the block 112.1 represents, for example, part of the block 112 explained with reference to FIG. 4.
- This block 112.1 comprises an algorithm for learning features of the candidates, for example in the form of a neural network or a so-called variational autoencoder.
- Data parameters (signals) and features selected by the user e.g. in the form of the parameters described above) (especially for the (time) domain) are used to train the algorithm.
- the characteristics of a respective candidate are represented, for example, by numerical values inserted in a one-dimensional or multi-dimensional vector space.
- a further machine learning model is created in block 112.2, e.g. in the form of a random forest classifier, a logistic regressor or a so-called support vector machine.
- This machine learning model is designed to predict whether a further (new, unclassified) data set has an anomaly or not.
- the block 112.2 represents, for example, part of the block 112 explained with reference to FIG. 4.
- the block 112.2 also receives the classification of the user via a communication link K6.
- the machine learning model represents a classification learning algorithm.
- the flow of information for training machine learning models takes place via the communication links K4 and K5 (dashed lines).
- the next candidate to be displayed to the user is determined by means of active learning.
- the trained algorithm for learning features is used for each candidate (for example each data record stored in the database 102) applied to identify characteristics of the candidates.
- the classification learning algorithm is used to determine which of the candidates is the one that promises the greatest learning effect (is the newest in terms of its characteristics, so to speak). For this purpose, for example, a distance to a separation boundary between different classes and / or to a separation boundary of a single class is determined.
- the maximum information content that is relevant for the machine learning model is determined. For example, an entropy value is determined.
- blocks 112.1 and 112.2 are connected via communication links K7, K8.
- FIG. 11A shows, by way of example, features of a multiplicity of candidates plotted in two dimensions (e.g. two different temperature or pressure values or the like). Each point plotted represents a candidate.
- the candidates are divided into two classes and the task of the classification learning algorithm (e.g. in the form of the machine learning model 51 or 57) is to find out which class the candidates belong to.
- the classes correspond to different states of the gas turbine engine 10 or, in general, of a machine. For example, one class indicates a defective state, the other class a correct state.
- Figure 11B shows the same candidates with some randomly selected candidates highlighted that have been classified and used to train the classification learning algorithm. Based on this, a separation line L separating the classes is drawn. This is suboptimal.
- FIG. 11 C shows selected candidates highlighted by means of active learning. These result in a separation line (e.g. determined using a regression algorithm), which enables a much better separation of the classes.
- active learning for example, the candidate who is closest to the separation line is always selected for this purpose. Alternatively or additionally, those candidates are selected who have the lowest confidence level in the Have prediction of the class.
- so-called entropy sampling is used. The most informative data sets are selected in this way. It should be mentioned that this also enables training that is not only particularly precise but also enables robust predictions to be made particularly quickly.
- the separation line can be curved or straight, and also be part of a multi-dimensional separation boundary. It is understood that the invention does not relate to those described above
- gas turbine engine 10 instead of the gas turbine engine 10, another machine, in particular generally a motor and / or power unit, for example a piston engine, can also be used.
- a motor and / or power unit for example a piston engine
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Mechanical Engineering (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| DE102020201239.3A DE102020201239A1 (de) | 2020-01-31 | 2020-01-31 | Sensorsystem und Verfahren zum Erkennen eines Zustands zumindest einer Maschine |
| PCT/EP2021/050910 WO2021151701A1 (de) | 2020-01-31 | 2021-01-18 | Sensorsystem und verfahren zum erkennen eines zustands zumindest einer maschine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| EP4097688A1 true EP4097688A1 (de) | 2022-12-07 |
Family
ID=74187316
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| EP21700767.3A Withdrawn EP4097688A1 (de) | 2020-01-31 | 2021-01-18 | Sensorsystem und verfahren zum erkennen eines zustands zumindest einer maschine |
Country Status (4)
| Country | Link |
|---|---|
| US (1) | US20230086842A1 (de) |
| EP (1) | EP4097688A1 (de) |
| DE (1) | DE102020201239A1 (de) |
| WO (1) | WO2021151701A1 (de) |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8165826B2 (en) * | 2008-09-30 | 2012-04-24 | The Boeing Company | Data driven method and system for predicting operational states of mechanical systems |
| JP5363927B2 (ja) | 2009-09-07 | 2013-12-11 | 株式会社日立製作所 | 異常検知・診断方法、異常検知・診断システム、及び異常検知・診断プログラム |
| US20170090866A1 (en) | 2015-09-25 | 2017-03-30 | Robert L. Vaughn | Universal sensor and/or sensor cluster to provide a detection pattern |
| US10401847B2 (en) * | 2016-12-12 | 2019-09-03 | General Electric Company | System and method for issue detection of industrial processes |
| US10599205B2 (en) * | 2017-09-18 | 2020-03-24 | Verizon Patent And Licensing Inc. | Methods and systems for managing machine learning involving mobile devices |
| US12045316B2 (en) * | 2018-03-09 | 2024-07-23 | Ciena Corporation | Automatic labeling of telecommunication network data to train supervised machine learning |
| US10969773B2 (en) * | 2018-03-13 | 2021-04-06 | Applied Materials, Inc. | Machine learning systems for monitoring of semiconductor processing |
-
2020
- 2020-01-31 DE DE102020201239.3A patent/DE102020201239A1/de active Pending
-
2021
- 2021-01-18 EP EP21700767.3A patent/EP4097688A1/de not_active Withdrawn
- 2021-01-18 US US17/795,992 patent/US20230086842A1/en not_active Abandoned
- 2021-01-18 WO PCT/EP2021/050910 patent/WO2021151701A1/de not_active Ceased
Non-Patent Citations (1)
| Title |
|---|
| RELAN RISHI: "Health Monitoring of Gas Turbine Engine Framework Design and Strategies Rishi Relan A Dissertation Submitted In Partial Fulfilment of the Requirements for the Degree of Master of Philosophy in the Department of Automatic Control and Systems Engineering", 1 June 2015 (2015-06-01), pages 1 - 175, XP093239406, Retrieved from the Internet <URL:https://etheses.whiterose.ac.uk/9576/1/MPhil_RishiRelan_110263933_Final_corrections.pdf> [retrieved on 20250114] * |
Also Published As
| Publication number | Publication date |
|---|---|
| US20230086842A1 (en) | 2023-03-23 |
| WO2021151701A1 (de) | 2021-08-05 |
| DE102020201239A1 (de) | 2021-08-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP4034955B1 (de) | Training von maschinenlernmodellen zur datengetriebenen entscheidungsfindung | |
| DE69010193T2 (de) | Überwachung. | |
| EP3671632B1 (de) | Bildbasierte wartungsvorhersage und detektion von fehlbedienungen | |
| EP3100122B1 (de) | Verfahren zum bereitstellen von zuverlässigen sensordaten | |
| DE102017220140B4 (de) | Abrufvorrichtung, Abrufverfahren und Abrufprogramm | |
| DE102010025851A1 (de) | Verfahren zum Training eines Systems zur Klassifikation eines Wälzlagerzustands sowie Verfahren zur Klassifikation eines Wälzlagerzustands und System zur Klassifikation eines Wälzlagerzustands | |
| DE102017124135A1 (de) | Maschinenfehlervorhersage basierend auf einer Analyse von periodischen Informationen in einem Signal | |
| DE10241746B4 (de) | Verfahren zur zyklischen Qualitätsbewertung und Prozessüberwachung bei periodischen Produktionsprozessen | |
| DE102020101008B3 (de) | Verfahren und vorrichtung zur maschinenüberwachung mit kontinuierlicher verbesserung einer vorausschauenden wartungsdatenbank | |
| EP3340250B1 (de) | Bauteilidentifikation bei der fehlerbehandlung von medizinischen geräten | |
| DE102019203149A1 (de) | Zeitreihen-datenanalysevorrichtung | |
| WO2021038079A1 (de) | Verfahren und vorrichtung zum analysieren eines ablaufprozesses | |
| DE102022111622A1 (de) | Computerprogramm und Verfahren zur verbesserten Asbestanalytik mit künstlicher Intelligenz | |
| EP4097688A1 (de) | Sensorsystem und verfahren zum erkennen eines zustands zumindest einer maschine | |
| EP3921810B1 (de) | Verfahren und vorrichtung zum automatisierten identifizieren eines produktfehlers eines produkts und/oder zum automatisierten identifizieren einer produktfehlerursache des produktfehlers | |
| DE102019214665A1 (de) | Training eines Maschinenlernmodells zur datengetriebenen Entscheidungsfindung | |
| DE60208415T2 (de) | Verfahren zur optimierung von testdaten | |
| EP1983450A2 (de) | Verfahren und Vorrichtung zur Ermittlung und Bereitstellung von Information zu einem Bild | |
| WO2020127285A1 (de) | Verfahren und vorrichtung zum bestimmen von segmenten in empfangenen zeitreihendaten einer systemkomponente | |
| EP4184350A1 (de) | Computer-implementiertes verfahren zum erkennen eines eingabemusters in mindestens einer zeitreihe einer mehrzahl von zeitreihen | |
| DE102020207062A1 (de) | Verfahren und System zum Training eines Maschinenlernmodells | |
| DE102024114462B3 (de) | Verfahren und System zur Bewertung einer technischen Anlage | |
| DE10220124A1 (de) | Verfahren zur Überwachung des Zustands einer Maschine | |
| EP3787229A1 (de) | Verfahren und vorrichtung zur automatischen auswahl von analyseketten zur merkmalsextraktion | |
| DE102024126903A1 (de) | Sortieranlagensteuerung |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: UNKNOWN |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE |
|
| PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
| 17P | Request for examination filed |
Effective date: 20220818 |
|
| AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
| DAV | Request for validation of the european patent (deleted) | ||
| DAX | Request for extension of the european patent (deleted) | ||
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: EXAMINATION IS IN PROGRESS |
|
| 17Q | First examination report despatched |
Effective date: 20250207 |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE APPLICATION HAS BEEN WITHDRAWN |
|
| 18W | Application withdrawn |
Effective date: 20250325 |