WO2022096525A1

WO2022096525A1 - Sensor system

Info

Publication number: WO2022096525A1
Application number: PCT/EP2021/080546
Authority: WO
Inventors: Dirk Mayer; Oliver Mey; Christian Schmidt; Samuel Klein; André Schneider
Original assignee: Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V.
Priority date: 2020-11-04
Filing date: 2021-11-03
Publication date: 2022-05-12
Also published as: DE102020213899A1

Abstract

Sensor system (10) for monitoring the state of a mechanical component or of an installation comprising one or more mechanical components, having the following features: a first sensor (12) that is designed to detect a vibration of the installation and/or of the one or more mechanical components and to output a first sensor signal (12s) according to the vibration; a second sensor (14) that is designed to detect an acoustic signal at the installation or the one or more mechanical components (20) and to output a second sensor signal (14s) according to the acoustic signal; and an evaluation device that is designed to derive a state signal for the installation according to the first and second sensor signal (14s) and according to a rotational speed signal.

Description

sensor system

description

Embodiments of the present invention relate to a sensor system for monitoring the condition of a mechanical component or plant. Preferred exemplary embodiments relate to the sensor system with a focus on sensing, the sensor system with a focus on a learning mode or on the sensor system with a focus on other combinatorial aspects. A further exemplary embodiment relates to a method for monitoring the condition of a plant with one or more mechanical components and to a corresponding computer program.

Mechanical systems, for example, are monitored using conventional vibration-based sensors (vibration-based CM systems). Vibration sensors, for example, are used here. A typical monitoring example of a vibration sensor is a wind turbine or the drive train of a locomotive. With vibration sensors, for example, gross errors such as imbalances can be detected. In the case of smaller or slower-running systems, monitoring with vibration sensors is difficult.

Based on this, there is a need for a concept for monitoring the condition of a mechanical component or system.

The object of the present invention is to create a concept for condition monitoring that offers an improved compromise between range of use, accuracy and configurability.

The object is solved by the subject matter of the independent patent claims.

Embodiments of the present invention create a sensor system for monitoring the condition of a mechanical component or a system with one or more mechanical components. The sensor system includes a first sensor, such as a vibration sensor, which is designed to detect a vibration of the system and / or the one or more mechanical components and depending ability to output a first sensor signal from the vibration. The sensor system also includes a second sensor, such as an active acoustic emission (AE) sensor or a passive acoustic emission sensor, which is designed to detect a high-frequency structure-borne noise signal on the system or the one or more mechanical components and to output a second sensor signal as a function of this signal. Acoustic emission is synonymous with acoustic emission or sound emission, for example in the sense of high-frequency structure-borne noise. Based on the first sensor signal and/or the second sensor signal, information about a state of the system or the mechanical component is possible, with the first or the second sensor signal being preferred in corresponding exemplary embodiments depending on the operating state. The selection can be determined, for example, using a speed signal. The main exemplary embodiment therefore includes an evaluation device which is designed to derive an additional signal for the system on the basis of the first and second sensor signals and a speed signal.

Embodiments of the present invention are based on the finding that by combining two sensors, which can determine, for example, structure-borne noise in different frequency ranges (vibration and acoustic emission (AE) in the ultrasonic range), a condition can be monitored much more precisely because, depending on the system or Operating state of the system one or the other sensor signal can be taken into account more. For example, the second sensor signal (e.g. based on ultrasound) can be weighted more heavily in the case of a slowly rotating system or a system in a slowly rotating state, while the first sensor signal (vibration signal) is given greater weight in fast-rotating systems or a system in a rapidly rotating state will. Therefore, according to exemplary embodiments, the speed signal is used during sensing in order to fuse the two sensor signals with one another, taking into account their speed dependency.

At this point it should be noted that a vibration is typically in the form of structure-borne noise and can have a frequency range of 0.1 Hz to 100 Hz, to 500 Hz, to 5 kHz, to 10 kHz or to 20 kHz, for example. A signal resulting from an acoustic emission event typically has a frequency range of > 100 Hz, > 500 Hz, > 5 kHz, > 10 kHz or > 20 kHz. Corresponding to a preferred variant > 20 kHz, ie outside the "audible range". An acoustic audio signal can be used instead of the acoustic audio signal. The acoustic audio signal can originate from an acoustic emission event and/or be in the range of more than 0.1 Hz, more than 0.1 kHz, more than 0.5 Hz, more than 0.5 kHz or more than 10 kHz.

According to exemplary embodiments, a third sensor is used to determine the rotational speed of the system or of the one or more mechanical components, which sensor detects the rotational speed and outputs the rotational speed signal as a function of the rotational speed. Alternatively, the evaluation device can be designed to determine the speed signal on the basis of the first and/or second sensor signal. In this respect, the evaluation device advantageously uses the speed signal, e.g. B. to weight or prefer the first and the second sensor signal to each other.

It is thus advantageously possible, by using or adding the speed signal, to open up broad classes of applications and to quickly adapt the measuring system to relevant phenomena. The evaluation of different frequency ranges also increases robustness by combining the analysis in different frequency ranges.

Therefore, the evaluation device includes a fuser, which is designed to fuse the first and the second sensor signal, possibly taking the speed signal into account. In accordance with exemplary embodiments, the fuser can be designed to fuse the first and the second sensor signal, taking into account a weighting that depends on the rotational speed. For example, the portion of the second sensor signal is larger, the lower the speed is. According to exemplary embodiments, the result of this merging and subsequent evaluation is a status signal which, for example, outputs maintenance information and/or indicates a need for maintenance.

According to one embodiment, the sensor signals (first and second) or the speed signal are recorded over time and evaluated and/or stored. According to exemplary embodiments, the evaluation device is designed to detect a change in the first and/or the second sensor signal and/or the speed signal over time. For example, the evaluation device can detect a change in the first and second sensor signals in pairs in each case over the respective time horizon. Corresponding to further exemplary embodiments it would be conceivable for the evaluation device to detect the change in the first and second sensor signals together with the speed signal, that is to say in a combination of all three signals, over time or the respective time horizon. Evaluation over time has the advantage that different speeds may be approached over time, with the combination of information for the different speed points being possible here as a result of the detection over time. When starting a system, for example, a low speed range is first approached and then a high speed range, so that, for example, in the high speed range, imbalances, misalignment, bending vibrations of the drive shaft or similar are detected, with damage to the bearing cage, defective rolling elements, missing lubrication in a low speed range and gearing damage can be seen. Structure-borne noise is preferably analyzed in the low speed range, while the ultrasonic signal is preferably evaluated in the high speed range. Generally speaking, different phenomena require measurements in different frequency ranges. It should be noted that the frequency range of the phenomenon is not necessarily related to speed. The frequency ranges vary depending on the size and drive train, but generally all depend on the speed.

According to exemplary embodiments, it would also be conceivable for the evaluation device to be designed to carry out a pre-classification of the first and/or the second sensor signal in order to output pre-classified first and/or second sensor signals. Advantageously, the indication effect or indication capability of the individual signals can be recognized with a pre-classification. According to preferred exemplary embodiments, there is a classification per sensor, which enables scalability. In the pre-classification, for example, the sensor signal is evaluated to determine whether relevant parts are present or not. This then also reduces the calculation effort for the general evaluation.

According to exemplary embodiments, the evaluation device can be trained with training data. For this purpose, the sensor system comprises an input interface, in which a damage event can be displayed externally to the evaluation device. Alternatively, another sensor can also be present, which directly detects the damage to the system or mechanical component. The evaluation device is designed, in particular during the learning mode, to store the damage event in combination with the recorded values of the first and/or the second sensor signal and in combination with the speed signal in order to use it for the system or a comparable Attachment to be used as a reference. During this training, it is evaluated in which sensor signal or in the combination of which sensor signals the damage event can be easily detected. The calculation effort is reduced when working with pre-classified data. According to further exemplary embodiments, the evaluation device is designed to store a damage event in combination with the values as learning data for a self-learning algorithm or a AI algorithm. According to exemplary embodiments, the self-learning algorithm or the KI algorithm is designed to recognize indicators for a status change in the status monitoring during execution. According to exemplary embodiments, the self-learning algorithm or AI algorithm can be designed to carry out status classifications according to the type and occurrence of a damage event. At least the first and second sensor signals and the speed signal are used here. According to further exemplary embodiments, the self-learning algorithm/AI algorithm is designed to weight the use of the first and second sensor signals relative to one another to determine an overall classification as a function of the rotational speed. In accordance with further exemplary embodiments, the self-learning algorithm/AI algorithm is designed to determine a characteristic curve for the weighting of the first and second sensor signals relative to one another as a function of a rotational speed. This characteristic can, for example, be designed in such a way that the first sensor signal is weighted more heavily than the second sensor signal at a lower speed, while the weighting changes accordingly with an increasing speed, so that the second sensor signal is then weighted more heavily than the first. According to further exemplary embodiments, the self-learning algorithm is designed to detect a sensor failure and, in this case, to carry out a status classification on the basis of the sensor signals that are still present.

According to one exemplary embodiment, the evaluation device is also designed in the learning mode to carry out a pre-classification of the first and/or second sensor signal, to output pre-classified first and/or second sensor signals, or to carry out a pre-classification for the first sensor signal and to carry out a pre-classification for the second sensor signal perform to output pre-classified first and / or second sensor signals. Alternatively, the evaluation device can be designed to execute a self-learning algorithm/K1 algorithm, which is designed to analyze pre-classified first and/or second sensor signals or the first and/or second sensor signals with regard to their indication capability, or to analyze the pre-classified first and/or or second sensor signals or the first and/or second sensor signals to analyze with regard to their indication capability in order to derive a weighting or selection of the respective first or second sensor signal for a merger (as intermediate levels of the class). Alternatively, the self-learning algorithm or Kl algorithm can be designed to analyze the preclassified first and/or second sensor signals or the first and/or second sensor signals in their relationship to the speed signal with regard to their indication capability, or to analyze the preclassified first and/or second To analyze sensor signals or the first and/or second sensor signals in relation to the speed signal with regard to their indication capability in order to derive a weighting or selection of the respective first or second sensor signal for a merger (as intermediate levels of Kl) taking into account the speed signal.

With regard to the mode of operation, it should be noted that the self-learning algorithm or AI algorithm is designed to randomly set the first and second sensor values or the speed value to zero or a predetermined value for the learning. In this way, it can advantageously be recognized whether the respective sensor value is currently meaningful. According to one embodiment, for the learning, the first and second sensor values are classified separately to train a combined classifier or to classify all sensor values together with the speed to train a combined classifier. Alternatively, the sensor values can be classified and trained together with the speed.

A further exemplary embodiment creates a sensor system in which an additional first and an additional second or also an additional sensor of a different type is provided. The evaluation function is then designed to take into account the additional first/additional second/additional other sensor in the evaluation or also in the training. Examples of sensors of another type are temperature sensors (thermography) or motor current sensors.

At this point it should be noted that, according to exemplary embodiments, each first and/or each second sensor can be arranged on a separate unit. Alternatively, the sensor system can also include only one unit, which has the first and the second sensor.

With regard to the system, it should be noted that, according to one embodiment, this is a rotary unit, such as a rotating shaft, at least one rotary Bearing or at least has a mechanical bearing. In this case, the one or more mechanical components can represent a shaft bearing or the shaft itself.

A further exemplary embodiment provides a method for monitoring the condition of a system with one or more mechanical components. The procedure includes the following steps:

- Detecting a vibration on the system or the one or more mechanical components by means of a first sensor and outputting a first sensor signal as a function of the vibration;

- Detecting an acoustic emission signal on the system or the one or more mechanical components with a second sensor and outputting a second sensor signal as a function of the acoustic emission signal; and

- Detection of the speed of the system; and

- Deriving a status signal for the system based on the first and/or second sensor signal and the speed of the system.

According to exemplary embodiments, the method can be computer-implemented, so that a computer program with a program code for carrying out the method explained above or at least one method step is created when the program runs on a computer.

Further formations are defined in the dependent claims. Exemplary embodiments of the present invention are explained with reference to the accompanying drawings. Show it:

Fig. 1 a is a schematic block diagram of a measurement system according to a

basic embodiment;

Fig. 1 b is a schematic representation to illustrate different

Frequency ranges for condition monitoring to explain exemplary embodiments;

ADJUSTED SHEET (RULE 91) ISA/EP 1c and 1d show a schematic representation of a measuring system in a specific application according to exemplary embodiments;

Fig. 1 e is a schematic representation to illustrate a measurement with the

Measuring system from FIG. 1 c for explaining exemplary embodiments;

1f to h schematic diagrams for the representation of possible test results for the explanation of exemplary embodiments;

2 shows a schematic block diagram of an evaluation logic associated with the measuring system according to exemplary embodiments; and

3 shows a schematic representation of an extended measurement system according to further exemplary embodiments.

Before exemplary embodiments of the present invention are explained below with reference to the figures, it should be noted that elements and structures that have the same effect are provided with the same reference symbols, so that the description of them can be applied to one another or are interchangeable.

1a shows a schematic block diagram of a measuring system 10 with a first sensor 12 and a second sensor 14. In addition, an optional third sensor 16 is illustrated. All sensors 12 to 16 are connected to an evaluation unit 18 .

The first sensor 12 is, for example, a vibration sensor that is designed to detect vibrations in a system to be monitored or a component 20 of a system to be monitored, here an axle 20, and to output a corresponding sensor signal 12s. The frequencies of this sensor are in the range of 0.1 Hz, 10 kHz or even 20 kHz, depending on the dimensioning. The vibration sensor 12 preferably measures vibrations or low-frequency structure-borne noise, which are, for example, in a frequency range from 0.1 to 100 Hz or from 0.5 to 500 Hz or generally 0.1 Hz to 20 kHz. Depending on the implementation, higher frequencies, e.g. B. up to 5 kHz, 10 kHz or 20 kHz can be measured, so that a measuring range of 0.1 Hz to 5/10/20 kHz is then set. In this respect, generally speaking, the sensor 12 can be designed to detect structure-borne noise in the audible range. rich to detect. The sensor 14 also detects structure-borne noise, but preferably in the higher frequency range, z. B. from 500 Hz or 1 kHz or from 10 kHz or 20 kHz, i.e. in the audible range or ultrasonic range (generally speaking, acoustic emissions). According to a preferred variant, acoustic emission (AE), for example in the ultrasonic range, ie >20 kHz (cf. above) can be measured. The sensor can be designed as a passive acoustic emission sensor, for example. According to preferred exemplary embodiments, the sensors 12 and 14 complement one another in such a way that their frequency ranges border one another or partially overlap in order to cover a very broad measurement range overall.

Based on the noise emission determined by component 20, sensor 14 emits a measurement signal 14s.

As shown in FIG. 1b, different damage events can be detected better in different frequency ranges. For example, a wind energy plant 20w is used to explain which damage events are easy to detect in which frequency range. The spectrum s20w of the wind turbine 20w extends from 0.1 to 25000 of any unit (e.g. Hertz). In the range from 0.1 to 1000, for example, resonances can occur in the drive train of the wind turbine. These indicate imbalance, misalignment, or flexural vibration in the driveshaft. The (geometry dependent) wavelength and frequency of a propagating (longitudinal/compression) vibration can _be related using the formula Csoiid = F, where Csoiid is the speed of sound in the material, F is the frequency and X is the wavelength. In the 20 to 25,000 range, other damage events, such as damage to the bearing cage, defective rolling elements, lack of lubrication or toothing damage, are easily recognizable. When using the sensor system 10 from FIG. 1a, a damage event in the low frequency range can be detected with the aid of the sensor 12 and a damage event associated with a high frequency range can be detected with the aid of the sensor 14. Frequently, not the entire spectrum measured with the individual sensor 12 or 14 is relevant for each damage event, but only individual frequencies.

Therefore, according to optional exemplary embodiments, the evaluation unit 18 is designed to classify each sensor signal 12s and 14s beforehand. Here, the evaluation is directed towards the respective frequency spectrum. The relevant spectrum Alternatively, the relevant frequency range can vary depending on the speed, as can be seen in the diagram from FIG. Also in order to be able to differentiate here, in the evaluation based on a rotational speed N, switching back and forth is made between the individual sensors 12 and 14 or preference is given to one of the sensor signals 12s or 14s. In this respect, the evaluation unit 18 is designed to select the sensor signals 12s or 14s or, preferably, to use them for the evaluation, based on the rotational speed N of the component 20, which is determined by means of the sensor 16. In addition, the rotational speed can also be taken into account within the sensor signal 12s or 14s in order to improve the evaluation of the signal 12s or 14s. According to one exemplary embodiment, the speed N can be taken into account when selecting the sensor signal 12s or 14s by weighting one sensor signal (1 from 12s + 14s) relative to the other sensor signal 12s or 14s as a function of the speed N. The weight can vary between 0 and 100%.

The consideration of different frequency ranges depending on the speed enables an adaptation to relevant phenomena on the one hand and on the other hand also a wide range of applications with the same sensor system 10. For example, the drive train of a locomotive 20I or the roller conveyor 20r can also be monitored using the same sensor system 10. The relevant range s20l of the drive train of the locomotive is, for example, between 0.8 and 30k of any unit (e.g. Hertz), while the relevant frequency range of the roller conveyor 20r, which is provided with the reference sign s20r, is between 5k and 50k. As can be seen, the high values, e.g. B. 20k to 50k can be assigned to the ultrasonic range, while for the low frequency ranges 0.1 to 1000 the relevant low range between 0.8 to 20k can be assigned to a vibration. The combination of the two sensors 12 and 14 to form a sensor system 10 increases the overall robustness of the analysis. Referring to FIGS. 1c, 1d and 1e, an exemplary application of the sensor system 10 to an electric motor 20e will now be explained. The electric motor 20e is equipped with a vibration sensor 12 at its output, for example. The output is connected to a shaft 20 which is monitored by an acoustic emission sensor 14. As can be seen from FIG. 1d, the sensor 14 is attached in the area of a roller bearing 20b. A speed sensor is not used here as the measurement signal, as in FIG Way on the electronics 21, a speed signal can be tapped. Electronics 21 also receives sensor signals 12s and 14s from sensors 12 and 14.

The current measurement result is shown in FIG. 1e. By means of the two sensors 12 and 14, the vibration, z. B. originating from an unbalanced mass (unbalanced mass 20m) determined. This signal is shown over time in diagram 12s + 14s. A pre-processing is carried out by the evaluation unit, so that the diagram appears for _2s + 14s. Since there is currently no balance weight, as shown by the value for the mass 20m, the result of the analysis then states that there is no damage event here. In order to generate the diagram 12s + 14s from the two sensor signals 12s and 14s, the operating parameter speed is also used here. Different damage events can be illustrated with this structure. In connection with FIGS. 1f to 1h, the pre-processing and detection of damage events will now be explained using three diagrams.

1f shows three acoustic emission signals plotted over time, which are made up as follows: a “burst type signal” 23b has individual pulses 23bp over time, which allow conclusions to be drawn about damage to components (here the rolling element). This "burst type signal" (burst signal) 23b is superimposed with a "continuous signal" (continuous noise signal, here the "background noise" of the drive train in operation) 23c in order to obtain the "mixed mode signal" (mixed signal) 23m, which corresponds to the measured sensor signal. A synthetic combination of this “mixed signal” 23m is shown in FIG. 1g. This is compared with a signal from FIG. 1f measured on the demonstrator. As can be seen, the signal curve is very similar to the synthetic signal. A so-called "lead break test" with a "brush" was carried out for the signal from 1 hour. The lead break test is carried out, for example, when the drive is stationary and generates a burst which approximately corresponds to signal a) from FIG. 1f).

The synthetic signal from FIG. 1g or the signal from FIG. 1f measured on the demonstrator can be used as a training data set for training data. According to exemplary embodiments, the evaluation unit 18, which can have a K1 algorithm or a self-learning algorithm, for example, is trained on the basis of a large number of training data sets. For example, several training data sets for different damage events or also diesel ben damage events of the evaluation unit 18 are fed, which then the characteristic points in the sensor data over time, such. B. characteristic frequencies or combinations of characteristic frequencies as a function of the speed. According to preferred exemplary embodiments, a multi-stage concept is used for this, which is explained with reference to FIG.

2 illustrates a multi-stage concept for evaluating sensor data. 2 shows an evaluation logic, here a KI evaluation logic 30, with a plurality of inputs 12'-17_n'. The evaluation logic 30 serves to classify and merge the sensor values of the signals from the inputs 12'-17_n' in order to output status information as a result.

The input for the sensor data is provided with the reference symbols 12', 14', 17 1 ', 17_n'. In addition, an additional input 16' is conceivable. It can be assumed, for example, that 12' represents the input for the vibration/structure-borne noise sensor, 14' represents the input for the acoustic sound sensor, 16' represents the input for the speed sensor, with additional sensors such as e.g. B. a temperature sensor or the like can be added.

Each sensor signal 12s, 17_1s and 17_ns is preclassified with a preclassification 31a-n. In other words, this means that a preclassifier 31a-n is provided for each sensor type, which can be adapted to the respective sensor type, for example. The provision of a pre-classifier 31a to 31n has the advantage that the entire system can be scaled as desired and can be extended to other applications with other sensor signals.

The pre-classified results 12sk, 14sk, 17 1sk and 17_nsk are combined in an intermediate level of the KI evaluation unit 30. This intermediate level is provided with the reference number 32 and prefers individual pre-classified signals 12sk to 17_nsk. The preference can be given, for example, on the basis of the input signal 16s of the additional input 16'. In addition, a combination of the individual signals is also carried out in the intermediate level, so that, for example, depending on the speed, switching back and forth between the two inputs 12' and 14' or these are weighted differently. The combined classification then takes place in the downstream classifier 34.

The combined classifier 34 aggregates the intermediate results and thus leads to greater accuracy through the use of additional input data. Furthermore, also Weaknesses and failures of individual sensors, e.g. B. compensated at certain speeds.

The result 35 is then a status result, so that, for example, based on the speed, it was recognized in the intermediate level that the sensors 12' are preferably used in combination with 17 1 for the classification, with the combination of values in the classifier 34 then showing that such values represent an A or B state. A state A can, for example, be a slightly encrypted state of the mechanical system.

According to exemplary embodiments, the sensor signal itself is evaluated in the pre-classification, so that the relevant measurement data, such as e.g. B. frequency ranges or general measurement ranges are filtered out and already pre-classified.

With regard to the classification process, it should be noted that different sensor data combinations, e.g. B. are fed to the individual inputs 12 'to 17n' from different types of damage that have occurred and these sensor data are evaluated after pre-classification. Additional information about the current state, e.g. B. a case of damage that has occurred is passed on, so that the latter can draw the following conclusions based on this additional information:

1. Recognition of the characteristic component in the sensor signal 12s to 17_ns using the pre-classification 12a to 17n.

2. Recognition of the relevant signals 12s to 17_ns by means of the intermediate level 32, for example to find out a weighting or a single characteristic signal from the plurality of signals 12s to 17ns and combining the preselected and preclassified signals to form the status information 35.

3. Assigning a damage event to a sensor value combination.

According to exemplary embodiments, the K1 algorithm is designed here, for example, to switch back and forth between the sensor signals for the state classification, a weighting, e.g. B. depending on the additional input 16 'to analyze and for to store subsequent measurements or to mutually validate the sensor signals based on the majority of the sensor signals. In other words, this means that the AI algorithm analyzes, verifies and evaluates the indication capability of the individual sensor signals, so that the informative value/indication capability of the sensor signals for the operation is known.

According to a preferred embodiment, the weighting takes place between the first sensor signal, e.g. B. the acoustic emission sensor, and the second sensor signal, z. B. the vibration sensor, depending on a speed. As already explained, for example, one sensor type is more meaningful at low speeds and the other sensor type at high speeds. In order to take these into account, the weighting is carried out between the two sensor signals or between signals derived from the sensor signals as a function of a speed signal. This speed dependency is not necessarily the same from system to system. For example, a small system may work preferentially with one type of sensor over a very narrow (low) speed range and work preferentially with the other type of sensor over a wide (high) speed range, while a substantially larger system may use a wide (low) speed range, for example of one type of sensor and a narrow (high) speed range with the other type of sensor. According to exemplary embodiments, there is not just one transition frequency, but rather transition frequencies that are gliding in all transition ranges. These are taken into account in that, for example, the weighting ratio of one sensor signal to the other sensor signal is varied as a function of the speed. This variation can be based on a characteristic of the weighting. In order to find out this characteristic curve, a self-learning algorithm or AI algorithm is used according to exemplary embodiments, which determines the corresponding weighting during operation and/or during a learning phase. This approach advantageously makes it possible for one and the same sensor system or, in particular, one and the same evaluation for the sensor system for a wide variety of systems (e.g. small, medium-sized or large) to be used and for the evaluation device or the evaluation algorithm to adapt itself to the corresponding system in a self-learning manner , by determining the optimal weighting characteristic. Two examples of weighting characteristics for two different system types are explained below. A first type of plant refers to a large rotary machine, such as. B. a wind turbine. Here, the acoustic emission sensor is preferably used in low speed ranges and the vibration sensor in high speed ranges. number ranges used. At low speeds, e.g. B. less than 10 revolutions per minute, then, for example, the weighting is 90:1. At high speeds (e.g. greater than 25 revolutions per minute), the weighting is then 10:90, for example. In the area in between, the weighting factor is varied accordingly. There can also be a weighting variation above and below the thresholds explained as an example, or any other dependency on the rotational speed, including a number of transition frequencies.

The second type of system is a roller of a roller conveyor. At high speeds, e.g. B. greater than 100 revolutions per minute, the focus is in particular on the vibration signal (90:1 weighting), while at lower speeds (less than 30 revolutions per minute) the focus is primarily on the acoustic emission signal (10:90). The area in between can in turn be viewed as a transition area using a characteristic curve.

According to exemplary embodiments, the characteristic is linear. According to further exemplary embodiments, the characteristic curve is progressive, S-shaped or has a different profile. The characteristic curve can also be designed in such a way that a sensor is no longer taken into account at all above a certain speed, so that a weighting factor of 0 then results here.

According to exemplary embodiments, it is not the sensor signal itself but the signals preprocessed with parts of the neural network that are weighted and fused. Hence the addition "or signals derived from the sensor signals". An exact value is not absolutely necessary, in the neural network it is typically values in the value range [0,1]. It is important to mention that it can also be more than one weight value. An example of this would be that, on the basis of the respective sensor values, a status classification according to three status classes (good, error 1, error 2) is carried out. For example, there can then be 3 weight values, where the confidence levels of the individual classifiers are weighted depending on the class.

Referring to FIG. 3, an extended exemplary embodiment of a sensor system 10' is explained. The sensor system 10' includes the data processing unit 20', which receives the information from a sensor platform 14* on the one hand and from a further sensor platform 12* and an optional further data source 17* on the other hand. Via the sensor platform 12* (DAQ front end), the rotational speed signal (cf. reference number 16), a vibration signal (cf. reference number 12) and a motor current signal (cf. reference number 16m) and made available to the evaluation unit 20 as sensor data.

The evaluation unit 20 can be used by the machine learning function 35, e.g. B. be expanded to an interface for receiving or unit for detecting damage events. The unit 20 is operated in combination with the machine learning 35 by a GUI 37.

Although some aspects have been described in the context of a device, it is understood that these aspects also represent a description of the corresponding method, so that a block or a component of a device is also to be understood as a corresponding method step or as a feature of a method step. Similarly, aspects described in connection with or as a method step also constitute a description of a corresponding block or detail or feature of a corresponding device. Some or all of the method steps may be performed by hardware apparatus (or using a hardware Apparatus), such as a microprocessor, a programmable computer, or an electronic circuit. In some embodiments, some or more of the essential process steps can be performed by such an apparatus.

Depending on particular implementation requirements, embodiments of the invention may be implemented in hardware or in software. Implementation can be performed using a digital storage medium such as a floppy disk, DVD, Blu-ray Disc, CD, ROM, PROM, EPROM, EEPROM or FLASH memory, hard disk or other magnetic or optical memory, on which electronically readable control signals are stored, which can interact with a programmable computer system in such a way or interact that the respective method is carried out. Therefore, the digital storage medium can be computer-readable.

Thus, some embodiments according to the invention comprise a data carrier having electronically readable control signals capable of interacting with a programmable computer system in such a way that one of the methods described herein is carried out. In general, embodiments of the present invention can be implemented as a computer program product with a program code, wherein the program code is operative to perform one of the methods when the computer program product runs on a computer.

The program code can also be stored on a machine-readable carrier, for example.

Other exemplary embodiments include the computer program for performing one of the methods described herein, the computer program being stored on a machine-readable carrier.

In other words, an exemplary embodiment of the method according to the invention is therefore a computer program that has a program code for performing one of the methods described herein when the computer program runs on a computer.

A further exemplary embodiment of the method according to the invention is therefore a data carrier (or a digital storage medium or a computer-readable medium) on which the computer program for carrying out one of the methods described herein is recorded. The data carrier, digital storage medium, or computer-readable medium is typically tangible and/or non-transitory.

A further exemplary embodiment of the method according to the invention is therefore a data stream or a sequence of signals which represents the computer program for carrying out one of the methods described herein. For example, the data stream or sequence of signals may be configured to be transferred over a data communication link, such as the Internet.

Another embodiment includes a processing device, such as a computer or programmable logic device, configured or adapted to perform any of the methods described herein. Another embodiment includes a computer on which the computer program for performing one of the methods described herein is installed.

A further exemplary embodiment according to the invention comprises a device or a system which is designed to transmit a computer program for carrying out at least one of the methods described herein to a recipient. The transmission can take place electronically or optically, for example. For example, the recipient may be a computer, mobile device, storage device, or similar device. The device or the system can, for example, comprise a file server for transmission of the computer program to the recipient.

In some embodiments, a programmable logic device (e.g., a field programmable gate array, an FPGA) may be used to perform some or all of the functionality of the methods described herein. In some embodiments, a field programmable gate array may cooperate with a microprocessor to perform any of the methods described herein. In general, in some embodiments, the methods are performed on the part of any hardware device. This can be hardware that can be used universally, such as a computer processor (CPU), or hardware that is specific to the method, such as an ASIC.

The devices described herein may be implemented, for example, using hardware apparatus, or using a computer, or using a combination of hardware apparatus and a computer.

The devices described herein, or any components of the devices described herein, may be implemented at least partially in hardware and/or in software (computer program).

The methods described herein may be implemented, for example, using hardware apparatus, or using a computer, or using a combination of hardware apparatus and a computer.

The methods described herein, or any components of the methods described herein, may be performed at least in part by hardware and/or by software. The embodiments described above are merely illustrative of the principles of the present invention. It is understood that modifications and variations to the arrangements and details described herein will occur to those skilled in the art. Therefore, it is intended that the invention be limited only by the scope of the following claims and not by the specific details presented in the description and explanation of the embodiments herein.

Claims

patent claims

1 . Sensor system (10) for monitoring the condition of a mechanical component or a system with one or more mechanical components, with the following features: a first sensor (12) which is designed to detect a vibration of the system and/or the one or more mechanical components detect and output a first sensor signal (12s) as a function of the vibration; a second sensor (14) which is designed to detect an acoustic audio signal and/or an acoustic emission signal on the system or the one or more mechanical components (20) and to detect it as a function of the acoustic audio signal and/or the acoustic emission signal to output a second sensor signal (14s); and an evaluation device which is designed to derive a status signal for the system on the basis of the first and second sensor signals (14s) and a speed signal; wherein the first sensor signal (12s) and the second sensor signal (14s) and/or signals derived from 12s and/or 14s are weighted for determining the status signal, taking into account a weighting; and wherein the weighting is dependent on the speed signal; and characterized in that the evaluation device has a self-learning algorithm or AI algorithm which is designed to determine a characteristic curve of the weighting of the first (12s) and second sensor signal (14s) relative to one another as a function of the speed signal.

2. Sensor system (10) according to claim 1, wherein the vibration is in the form of a low-frequency structure-borne noise and/or wherein the vibration is in the range of 0.1 Hz to 20 kHz or 0.5 Hz to 10 kHz or 0.1 Hz to 10 kHz located.

3. Sensor system (10) according to any one of the preceding claims, wherein the acoustic audio signal originates from an acoustic emission event and/or in the range larger 0.1 Hz, greater than 0.1 kHz, greater than 0.5 Hz, greater than 0.5 kHz or greater than 10 kHz; and/or wherein the acoustic emission signal results from an acoustic emission event and/or is in the range greater than 10 kHz or 20 kHz.

4. Sensor system (10) according to one of the preceding claims, wherein the evaluation unit (18) is designed to detect the first and / or the second sensor signal (14s) and / or speed signal over time and evaluate and / or store.

5. Sensor system (10) according to one of the preceding claims, which comprises a third sensor which is designed to detect the current speed (n) of the system or the one or more mechanical components (20) and, depending on the speed (n ) output the speed signal; or wherein the evaluation device is designed to determine the speed signal on the basis of the first and/or second sensor signal (14s).

6. Sensor system (10) according to one of the preceding claims, wherein the evaluation device is designed to analyze the first and/or second sensor signal (14s) with regard to absolute values of the first and/or second sensor signal (14s) and/or with regard to evaluate changes in value of the first and/or second sensor signal (14s); and/or wherein the evaluation device is designed to detect a change in the first and/or second sensor signal (14s) and/or the speed signal over time; and/or wherein the evaluation device is designed to detect a change in the first and second sensor signal (14s) in pairs in each case over the respective time horizon; and/or wherein the evaluation device is designed to detect a change in the first (12s) and second sensor signal (14s) and the speed signal in combination of all three signals over time. Sensor system (10) according to one of the preceding claims, wherein the evaluation device is designed to carry out a pre-classification (31 a-31 n) of the first (12s) and/or second sensor signal (14s) in order to pre-classify first (12s) and/or to output second sensor signals (14s); or wherein the evaluation device is designed to carry out a pre-classification (31 a-31 n) for that of the first sensor signal (12s) and to carry out a pre-classification (31 a-31 n) for that of the second sensor signal (14s) in order to obtain pre-classified first (12s) and/or output second sensor signals (14s). Sensor system (10) according to one of the preceding claims, wherein the evaluation device has a fuser (34) which is designed to fuse the first (12s) and the second sensor signal (14s); or wherein the evaluation device has a fuser (34) which is designed to fuse the first sensor signal (12s) and the second sensor signal (14s) taking into account the speed signal; or wherein the evaluation device has a fuser (34) which is designed to fuse the first sensor signal (12s) and the second sensor signal (14s) taking into account the weighting which is dependent on the speed signal; or wherein the evaluation device has a fuser (34) which is designed to fuse the first sensor signal (12s) and the second sensor signal (14s) taking into account the weighting which is dependent on the speed signal; wherein the proportion of the second sensor signal (14s) is greater, the lower the speed signal is. Sensor system (10) according to one of the preceding claims, wherein the evaluation device is designed to derive maintenance information and/or a maintenance requirement as a status signal. Sensor system (10) according to any one of the preceding claims, wherein the sensor system (10) has an input interface via which a damage event Evaluation device can be displayed externally (for training purposes); and/or wherein the evaluation device comprises a further sensor (12) which detects damage to the system and/or the mechanical component (20); and/or wherein the evaluation device is designed to store a damage event in combination with recorded values of the first and/or second signal and/or the speed signal in order to use this as a reference for the system or a comparable system; and/or wherein the evaluation device is designed to store a damage event in combination with values of the first and/or second sensor signal (14s) and/or the speed signal as learning data for a self-learning algorithm or a KI algorithm. Sensor system (10) according to one of the preceding claims, wherein the evaluation device is designed to execute a self-learning algorithm and/or a KI algorithm in order to identify indicators for a change in status in the status monitoring; and/or wherein the self-learning algorithm or the AI algorithm is designed to carry out a state classification according to the type and occurrence of a damage event; and/or wherein the self-learning algorithm or the AI algorithm is designed to carry out a state classification on the basis of the first sensor signal (12s) and/or the second sensor signal (14s) and the speed signal; and/or wherein the self-learning algorithm or the AI algorithm is designed to weight the use of the first (12s) and second sensor signal (14s) relative to one another to determine an overall classification as a function of the speed signal; ; and or wherein the self-learning algorithm or the AI algorithm is designed to detect a sensor failure and, in this case, to carry out a status classification on the basis of the sensor signals that are still valid.

12. Sensor system (10) according to one of the preceding claims, wherein the evaluation device is designed to carry out a pre-classification (31 a-31 n) of the first (12s) and/or second sensor signal (14s) in order to pre-classify first (12s) and /or to output second sensor signals (14s); or to carry out a pre-classification (31 a-31 n) for the first sensor signal (12s) and to carry out a pre-classification (31 a-31 n) for the second sensor signal (14s), to pre-classify first (12s) and/or second sensor signals (14s) to output; and/or wherein the evaluation device is designed to execute a self-learning algorithm and/or a KI algorithm, the self-learning algorithm or KI algorithm being designed to use the pre-classified first (12s) and/or second sensor signals (14s) or the first (12s) and/or second sensor signals (14s) with regard to their indication capability, or to analyze the pre-classified first (12s) and/or second sensor signals (14s) or the first (12s) and/or second sensor signals (14s) with regard to their indication capability to analyze in order to derive a weighting or selection of the respective first (12s) or second sensor signal (14s) for a fusion (as intermediate levels of the KI); or wherein the evaluation device is designed to execute a self-learning algorithm and/or a KI algorithm, wherein the self-learning algorithm or KI algorithm is designed to use the pre-classified first (12s) and/or second sensor signals (14s) or the first (12s ) and/or second sensor signals (14s) in relation to the speed signal with regard to their indication capability, or to analyze the pre-classified first (12s) and/or second sensor signals (14s) or the first (12s) and/or second sensor signals (14s ) in their relation to the speed signal with regard to their indication capability in order to derive a weighting or selection of the respective first (12s) or second sensor signal (14s) for a merger (as intermediate levels of Kl) taking into account the speed signal.

13. Sensor system (10) according to one of claims 1 1 or 12, wherein the self-learning algorithm and/or the AI algorithm is designed to randomly set the first or second sensor values or the speed value to zero or a predetermined value for the learning; and/or wherein the self-learning algorithm and/or the KI algorithm is designed to classify the first or second sensor values separately for the learning in order to train a combined classifier, or to classify all sensor values together with the speed value to form a combined train classifier; or classify and/or train the first or second sensor values along with the speed value.

14. Sensor system (10) according to one of the preceding claims, wherein the sensor system (10) has at least one additional first and/or one additional second sensor (14) and/or one additional sensor (17 1 ′) of a different type; and/or wherein the evaluation function is designed to also take into account the additional first and/or the additional second sensor (14) and/or the additional sensor (17 1 ') of a different type in the evaluation and in the training.

15. Sensor system (10) according to any one of the preceding claims, wherein each first (12) and / or each second sensor (14) is arranged on a separate unit.

16. Sensor system (10) according to any one of the preceding claims, wherein the sensor system (10) comprises a unit having the first sensor (12) and the second sensor (14); and/or wherein the sensor system (10) comprises a plurality of units each having a first (12) and a second sensor (14).

17. Sensor system (10) according to any one of the preceding claims, wherein the system has a rotation unit or a rotary bearing or a mechanical bearing.

18. Sensor system (10) according to claim 17, wherein the one or more components have a shaft and/or a bearing of a shaft. Sensor system (10) according to one of the preceding claims, wherein the evaluation device is designed to weight the second measured values compared to the first measured values depending on the rotational speed (n) or speed of the system, the weighting being determined by a Kl algorithm. Method for condition monitoring of a plant with one or more mechanical components (20), with the following steps:

detecting a vibration on the system or the one or more mechanical components by means of a first sensor (12) and outputting a first sensor signal (12') as a function of the vibration;

detecting an acoustic signal on the system or the one or more mechanical components with a second sensor (14) and outputting a second sensor signal (14') as a function of the acoustic signal; and

detecting the speed (n) of the system; and

calculation of derived signals (12s') and (14s') from the sensor signals 12s and 14s; and

Weighting the derived first sensor signal (12s') and the second derived sensor signal (14s') taking into account a weighting, the weighting being dependent on the speed signal;

deriving a status signal for the plant based on the first (12') and/or second sensor signal (14') and the speed (n) of the plant; characterized in that a self-learning algorithm or Kl algorithm is used to determine a characteristic curve of the weighting of the first (12s)/(12s') and second sensor signal (14s)/(14s') relative to one another as a function of the speed signal. Computer program with a program code for carrying out the method according to Claim 20, when the program runs on a computer.