SENSOR FOR DETECTING AT LEAST ONE PROPERTY OF A FLUID MEDIUM IN AT LEAST ONE MEASURING SPACE

Description

State of the art

[0001] A variety of sensors, sensor elements, and methods for detecting at least one property of a fluid medium in a measurement space are known from the prior art. These can, in principle, be any properties of a gaseous or liquid fluid medium, with one or more properties being detectable. The invention is described below, without limiting further embodiments and applications, particularly with reference to sensor elements for detecting a gas, in particular an H2 content in a measurement gas.

[0002] Sensor elements of the type described here are used in a wide variety of fields, for example in automotive engineering, process engineering, chemistry, and mechanical engineering, particularly for determining gas concentrations. For example, the determination of hydrogen concentrations, for example in an air-hydrogen mixture, plays a major role in the use of hydrogen fuel cell systems. Safety-relevant applications are also worth mentioning here. An air-hydrogen mixture becomes ignitable at a hydrogen content of around 4%. Sensor elements for detecting hydrogen can be used, for example, in hydrogen fuel cell vehicles to detect hydrogen escaping due to damage or a defect and, by coupling them to appropriate systems, to trigger warning signals and / or protective measures.Therefore, several hydrogen sensors are required per fuel cell vehicle, which are either installed in the exhaust system (exhaust) or operate under atmospheric conditions (ambient).

[0003] A variety of measurement principles can be used for such hydrogen sensors. These include, among others, thermal conduction, catalytic pellistor, electrochemical cell, semiconducting metal oxide, chemiresistor, and field-effect transistor.

[0004] Despite the advantages of state-of-the-art sensor elements for detecting at least one property of a fluid medium, they still have potential for improvement. For use in automotive technology, such a hydrogen sensor must meet certain requirements. The aforementioned measurement principles each have certain deficiencies or disadvantages with regard to these requirements. Such sensor elements predominantly exhibit insufficient response time, a measurement range above the minimum measurement range, and / or cross-sensitivity to other components such as helium or volatile organic compounds. Furthermore, some of them are based on expensive assembly and connection technology.

[0005] From US 2013 / 311108 A1, a sensor for detecting at least one property of a fluid medium in at least one measuring chamber is known, which sensor has a first sensor element for detecting a thermal conductivity of the fluid medium and for outputting a first measurement signal, a second sensor element which has a semiconducting metal oxide and which is designed to output a second measurement signal, and a third sensor element for detecting a physical property of the fluid medium, wherein the third sensor element differs from the first sensor element and the second sensor element with regard to the detected physical property and is designed to output a third measurement signal.Furthermore, electronic means are provided for evaluating the first measurement signal, the second measurement signal and the third measurement signal and for outputting control signals for changing operating parameters of the first sensor element, the second sensor element and / or the third sensor element by means of an algorithm.

[0006] The technical article L. Boon-Brett et al: "Identifying performance gaps in hydrogen safety sensor technology for automotive and stationary applications", published in the International Journal of Hydrogen Energy 35 (2010), pages 373-384, discloses a multi-type sensor system consisting of MOx sensors and sensors based on the change in thermal conductivity, as well as a third sensor in the form of a pressure sensor.

[0007] EP 1 621 882 A2 discloses a combination of a metal oxide sensor, a thermal conductivity sensor and a temperature sensor.

[0008] EP 3 540 420 A1 discloses a sensor unit consisting of a TCD, a metal oxide sensor and a temperature sensor. Disclosure of the invention

[0009] Within the scope of the present invention, a sensor for detecting at least one property of a fluid medium in a measuring space is proposed, which at least largely avoids the disadvantages of known sensors for detecting at least one property of a fluid medium in a measuring space, and which offers sufficient sensitivity, measuring range, response time and selectivity with regard to the requirements in automotive technology.

[0010] A sensor according to the invention for detecting at least one property of a fluid medium in at least one measuring chamber, in particular for detecting an H 2 content in a measuring gas, comprises at least: a first sensor element configured to detect a thermal conductivity of the fluid medium and to output a first measurement signal, a second sensor element comprising a semiconducting metal oxide and configured to output a second measurement signal, a third sensor element configured to detect a physical property of the fluid medium, wherein the third sensor element differs from the first sensor element and the second sensor element with regard to the detected physical property and is configured to output a third measurement signal, and an electronic evaluation unit configured to evaluate the first measurement signal, the second measurement signal, and the third measurement signal, wherein the electronic evaluation unit is further configured to output control signals to the first sensor element, the second sensor element, and the third sensor element and to change operating parameters of the first sensor element,of the second sensor element and / or the third sensor element is implemented by means of an algorithm. According to the invention, the algorithm comprises an artificial neural network, wherein the electronic evaluation unit is configured to output correction signals to the first sensor element, the second sensor element, and the third sensor element for changing the operating parameters.

[0011] The information about the property of the fluid medium to be measured, such as the H2 concentration in the supplied measurement gas, is mainly generated by the first sensor element in the form of a thermal conductivity sensor element and the second sensor element in the form of an MOX sensor element (MOX = metal oxide). The thermal conductivity of a gas is inversely proportional to the square root of the mass of the gas molecules, so that gases with light atoms such as H2 molecules or He atoms have a significantly higher thermal conductivity than air, which essentially consists of N2 and O2 molecules. The higher the measured thermal conductivity, the greater the proportion of light molecules. The MOX sensor element consists of a semiconducting metal oxide such as SnO2 or WO3, the electrical resistance of which decreases when a chemically reducing gas such as hydrogen, methane or water vapor is contained in the air.The content of reducing gases in the air can be determined via the measured electrical resistance. However, the signals from the thermal conductivity and MOX sensor elements are not clearly determined by the hydrogen concentration in the air, because other gases in the air can also lead to the same measurement result. This is referred to as cross-sensitivity to other gases, which limits the absolute sensor accuracy. In order to guarantee the desired accuracy of 0.1% vol. for H2 for all conceivable air compositions, corrections in the signal processing are necessary by recording and evaluating additional measured variables. According to the invention, at least one further sensor element is therefore proposed for recording a further physical property of the fluid medium.For example, additional sensor elements for relative humidity, gas temperature, and gas pressure are integrated into the H2 sensor, whose signals are taken into account in the electronic evaluation unit, which serves as the central electronic signal processing unit. However, this is not done in the form of a characteristic curve map, but rather through machine learning methods such as training a neural network.

[0012] According to the invention, the electronic evaluation unit is designed to change operating parameters of the first sensor element, the second sensor element and / or the third sensor element by means of an algorithm, wherein the algorithm comprises an artificial neural network.

[0013] The neural network is trained before the sensor is delivered. The invention utilizes transfer learning methods, meaning the complete, time-consuming training of the neural network is performed on only a few product sensors. The remaining sensor products receive these node parameters during programming and only undergo brief training for the purpose of individual fine-tuning. Training involves successively exposing the sensor to air with different proportions of gases, primarily H2, or interfering gases such as He or CH4, at different relative humidities, air pressures, and temperatures. The set parameters for gas concentrations, relative humidity, air pressure, and temperature are used as training data for the neural network.

[0014] The third sensor element can be designed to detect at least one physical property selected from the group consisting of: humidity, in particular relative air humidity, pressure, in particular air pressure, and temperature, in particular air temperature.

[0015] It is explicitly emphasized that the sensor can have more than three sensor elements, for example four, five or more sensor elements.

[0016] As part of the complete training of the neural network, the operating parameters of the individual sensor elements are optimized, so that the time until the measured value is available and the error between the measured and actual H2 concentration in air are minimized. According to the invention, the temperature of the MOX element, which is typically heated for operation, or the temperature of the measuring element of the thermal conductivity sensor element is selected as the parameter to be trained.

[0017] The first sensor element, the second sensor element, and the third sensor element can be separate sensor elements. Alternatively, the first sensor element, the second sensor element, and the third sensor element can be integrated sensor elements in a single sensor chip.

[0018] Accordingly, the sensor elements do not necessarily have to be physically separate components connected by a circuit carrier. Sensor functions can also be integrated into a single component or chip. For example, the thermal conductivity sensor element and the MOX sensor element can be integrated into a single chip or module. Likewise, the humidity, pressure, and temperature sensor elements can be integrated into a single module or chip. All conceivable aggregations of individual or multiple functions in a single module are preferred.

[0019] The sensor may further comprise a voltage converter. The voltage converter may be configured for connection to an external voltage source. The voltage converter may further be configured to generate a supply voltage for the first sensor element, the second sensor element, and the third sensor element.

[0020] The voltage converter thus receives the supply voltage for the entire sensor from the outside and generates the supply voltages required for the sensor elements.

[0021] The sensor may further comprise an interface, wherein the interface is configured to receive control commands from an external control unit and / or to output measurement data from the sensor to an external control unit. The interface is, for example, a data exchange module. The data exchange module receives control commands for the entire sensor from the outside and outputs measurement data on the hydrogen concentration and other requested measurement data or metadata. Such a data exchange module is also known as a communication chip.

[0022] The sensor may further comprise a sensor housing. The first sensor element, the second sensor element, the third sensor element, and the electronic evaluation unit may be arranged in the sensor housing. The sensor housing may have at least one opening through which the first sensor element, the second sensor element, and the third sensor element can be exposed to the fluid medium.

[0023] The gas whose hydrogen concentration is to be measured is thus introduced into the housing via a gas-permeable opening in the sensor housing and fed to the sensor elements.

[0024] The electronic evaluation unit is designed to output control and correction signals to the first sensor element, the second sensor element, the third sensor element in order to change the operating parameters.

[0025] The sensor thus comprises several sensor elements and a central electronic processing unit in a sensor housing. The central electronic processing unit processes the signals from the sensor elements using machine learning methods to determine a measured value for the existing H2 concentration within < 1 s with a low error of <0.1% vol H2, which is then made available to the user via the sensor's output interface. It is also intended that, as part of the signal processing, the central electronic processing unit also sends control or correction signals to the sensor elements in order to improve the overall sensor performance or optimize the accuracy or response time.

[0026] In the context of the present invention, a sensor is basically understood to be any device which can detect at least one property of the fluid medium and which can, for example, generate at least one measurement signal corresponding to the detected property, for example an electrical measurement signal such as a voltage or a current. The measurement signal can be output by one or more sensor elements as components of the sensor. The property can, for example, be a physical and / or a chemical property. Combinations of properties can also be detectable. In particular, the sensor can be designed to detect at least one property of a gas, in particular an H2 content in a measurement gas. Other properties and / or combinations of properties can also be detectable.

[0027] The sensor can be configured, in particular, for use in a hydrogen fuel cell vehicle. The measuring chamber can, in principle, be any open or closed chamber in which the fluid medium, in particular the measuring gas, is contained and / or through which the fluid medium, in particular the measuring gas, flows.

[0028] In the context of the present invention, a housing is understood to mean any component or group of components that can completely or partially enclose the sensor element and / or seal it off from the outside and impart mechanical stability to the sensor element. In particular, a housing can enclose at least one interior space. For example, the housing can at least partially enclose the interior space and at least partially delimit it from its surroundings. In particular, the housing can be made entirely or partially from at least one of the following materials: a plastic; a metal, a ceramic, or a glass. Short description of the drawings

[0029] Further optional details and features of the invention emerge from the following description of preferred embodiments, which are shown schematically in the figures.

[0030] They show: Figure 1 a schematic representation of a sensor according to the invention for detecting at least one property of a fluid medium in at least one measuring space and Figure 2 a schematic representation of the electronic evaluation unit. Embodiments of the invention

[0031] Figure 1shows a schematic representation of a sensor 10 according to the invention for detecting at least one property of a fluid medium 12 in at least one measuring chamber 14, in particular for detecting an H2 content in a measuring gas 16. The sensor 10 can be designed in particular for use in a hydrogen fuel cell vehicle. However, other applications are also possible. The sensor 10 can in particular comprise one or more further functional elements not shown in the figures, such as electrodes, electrode leads and contacts, multiple layers or other elements. Accordingly, the sensor 10 can be installed in the exhaust system of the hydrogen fuel cell vehicle (exhaust) or operate under atmospheric conditions (ambient). Consequently, the measuring chamber can be an exhaust system or the interior of the hydrogen fuel cell vehicle.

[0032] The sensor 10 has a first sensor element 18. The first sensor element 18 is designed to detect a thermal conductivity of the fluid medium 12. The first sensor element 18 is also designed to output a first measurement signal.

[0033] The sensor 10 further comprises a second sensor element 20. The second sensor element 20 is a MOX sensor element. Thus, the second sensor element 20 comprises a semiconducting metal oxide. The second sensor element 20 is further configured to output a second measurement signal.

[0034] The sensor 10 further comprises a third sensor element 22. The third sensor element 22 is designed to detect a physical property of the fluid medium. The third sensor element 22 differs from the first sensor element 18 and the second sensor element 20 with regard to the detected physical property. In the embodiment shown, the third sensor element 22 is designed to detect humidity of the fluid medium 12. For example, the third sensor element 22 is designed to detect relative air humidity. The third sensor element 22 is designed to output a third measurement signal.

[0035] In the embodiment shown, the sensor 10 has more than three sensor elements. Thus, the sensor 10 further has a fourth sensor element 24. The fourth sensor element 24 is designed to detect a physical property of the fluid medium. The fourth sensor element 24 differs from the first sensor element 18 and the second sensor element 20 with regard to the detected physical property. In the embodiment shown, the fourth sensor element 24 is designed to detect a temperature of the fluid medium 12. For example, the fourth sensor element 24 is designed to detect an air temperature. The fourth sensor element 24 is designed to output a fourth measurement signal.

[0036] The sensor 10 further comprises a fifth sensor element 26. The fifth sensor element 26 is designed to detect a physical property of the fluid medium. The fifth sensor element 26 differs from the first sensor element 18 and the second sensor element 20 with regard to the detected physical property. In the embodiment shown, the fifth sensor element 26 is designed to detect a pressure of the fluid medium 12. For example, the fifth sensor element 26 is designed to detect an air pressure. The fifth sensor element 26 is designed to output a fifth measurement signal.

[0037] The sensor elements 18, 20, 22, 24, 26 can be separate sensor elements. Preferably, the sensor elements 18, 20, 22, 24, 26 are integrated into a sensor chip or sensor module (not shown in detail).

[0038] The sensor 10 further comprises an electronic evaluation unit 28. The electronic evaluation unit 28 is designed to evaluate the first measurement signal, the second measurement signal, and the third measurement signal. Furthermore, the electronic evaluation unit 28 is designed to evaluate the fourth measurement signal and the fifth measurement signal. For this purpose, the electronic evaluation unit 28 communicates with the sensor elements 18, 20, 22, 24, 26. The electronic evaluation unit 28 is further designed to change operating parameters of the first sensor element 18, the second sensor element 20, and / or the third sensor element 24. Furthermore, the electronic evaluation unit 28 is designed to change operating parameters of the fourth sensor element 24 and the fifth sensor element 26.The electronic evaluation unit 28 is configured to change the operating parameters of the first sensor element 18, the second sensor element 20, the third sensor element 22, the fourth sensor element 24, and the fifth sensor element 26 using an algorithm 30. The algorithm 30 comprises an artificial neural network. Machine learning methods are used for the artificial neural network. The electronic evaluation unit 28 is configured to output control and / or correction signals to the sensor elements 18, 20, 22, 24, 26 in order to change or adapt their operating parameters.

[0039] The sensor 10 further includes a voltage converter 32. The voltage converter 32 is designed for connection to an external voltage source (not shown in detail). The voltage converter 32 is also designed to generate a supply voltage for the sensor elements 18, 20, 22, 24, 26. In other words, the voltage converter 32 supplies the required supply voltage to the sensor elements 18, 20, 22, 24, 26.

[0040] Sensor 10 further includes an interface 34. Interface 34 is configured to receive control commands from an external control unit (not shown in detail). Interface 34 is also configured to output measurement data from sensor 10 to the external control unit. Accordingly, interface 34 can be configured as a data exchange module.

[0041] The sensor 10 further comprises a sensor housing 36. The sensor elements 18, 20, 22, 24, 26 and the electronic evaluation unit 28 are arranged in the sensor housing 36. Furthermore, the voltage converter 32 and the interface 34 are at least partially arranged in the sensor housing 36. The sensor housing 36 has at least one opening 38. The sensor elements 18, 20, 22, 24, 26 can be exposed to the fluid medium and contacted by it via the opening 38.

[0042] Figure 2 shows a schematic representation of the electronic evaluation unit 28. As already mentioned, an algorithm 30, which includes an artificial neural network, runs in the electronic evaluation unit. Figure 2shows in particular a schematic representation of the algorithm 30. The algorithm 30 is supplied with first measurement data 40 from the first sensor element 18, second measurement data 42 from the second sensor element 20, third measurement data 44 from the third sensor element 22, fourth measurement data 46 from the fourth sensor element 24 and fifth measurement data 48 from the fifth sensor element 46. By means of the algorithm 30, the H 2 concentration 50 in, for example, vol.% and the percentage achievement of the explosion limit 52 in percent of H 2 in air are determined from the measurement data 40, 42, 44, 46, 48.

[0043] The operation of sensor 10 is described below. In addition to the sensor elements 18, 20, 22, 24, 26 and the electronic evaluation unit 28 as the central electronic signal processing unit, the voltage converter 32 and the interface 34 as a data exchange module are included in the sensor 10. The voltage converter 32 receives the supply voltage for the entire sensor 10 from the outside and generates the supply voltage required for the sensor elements 18, 20, 22, 24, 26 from this. The data exchange module 34 receives control commands for the entire sensor 10 from the outside and outputs measurement data on the hydrogen concentration and other requested measurement data or metadata to the outside. The measurement gas 16, whose hydrogen concentration is to be measured, is introduced into the sensor housing 36 via the gas-permeable opening 38 in the sensor housing 36 and fed there to the sensor elements 18, 20, 22, 24, 26.

[0044] The information about the H2 concentration to be measured in the supplied measurement gas 16 is mainly generated by the first sensor element 18 as a thermal conductivity sensor element and the second sensor element 20 as an MOX sensor element. The thermal conductivity of a gas is inversely proportional to the square root of the mass of the gas molecules, so that gases with light atoms such as H2 molecules or He atoms have significantly higher thermal conductivity than air, which essentially consists of N2 and O2 molecules. The higher the measured thermal conductivity, the greater the proportion of light molecules. The second sensor element 20 has or consists of a semiconducting metal oxide such as SnO2 or WO3, the electrical resistance of which decreases when a chemically reducing gas such as hydrogen, methane or water vapor is contained in the air. The measured electrical resistance can be used to determine the content of reducing gases in the air.As described, the signals of the first sensor element 18 and the second sensor element 20 are not uniquely determined by the hydrogen concentration in the air, as other gases in the air can also lead to the same measurement result. This is referred to as cross-sensitivity with regard to other gases, which limits the absolute sensor accuracy. In order to guarantee the desired accuracy of 0.1% vol. for H 2 for all conceivable air compositions, corrections in the signal processing are necessary by recording and evaluating additional measured variables. According to the invention, sensor elements 22, 24, 26 for the relative humidity, the gas temperature and the gas pressure are integrated in the H 2 sensor, and their signals are taken into account in the electronic evaluation unit 28. This consideration, however, does not take place in the form of a characteristic field, but rather through machine learning methods such as, for example,by training a neural network of algorithm 30 as shown schematically in . Fig. 2 is shown. The training of the neural network takes place before delivery of the sensor 10. The methods of transfer learning are used in such a way that the complete time-consuming training of the neural network is only carried out on a few product sensors. The other sensor products receive these node parameters during programming and only undergo a short training session for the purpose of individual fine-tuning. The training consists of sequentially exposing the sensor 10 to air with different proportions of gases, primarily H2, or interfering gases such as He or CH4, at different relative humidities and different air pressures and temperatures, and using the set parameters for the gas concentrations, relative humidity, air pressure and temperature as training data for the neural network.

[0045] As part of the complete training of the neural network, the operating parameters of the individual sensor elements 18, 20, 22, 24, 26 are optimized according to the invention, so that the time until the measured value is available and the error between the measured and actual H2 concentration in air are minimized. According to the invention, the temperature of the second sensor element 20, which is typically heated for operation, or the temperature of the measuring element of the first sensor element 18 is selected as the parameter to be trained.

[0046] The sensor 10 according to the invention is characterized by the presence of a sensor element for thermal conductivity and a sensor element with a semiconducting metal oxide, as well as at least one sensor element for another physical quantity, such as relative humidity, pressure, and temperature, as well as a central electronic signal processing unit. Furthermore, the invention is recognizable in the product in operation by the fact that the operating parameters for the sensor elements adapt to the external conditions and thus always meet the requirements for measurement accuracy for H 2 -% vol. and response time. Circuit elements or units optimized for neural networks can be recognized in the central electronic signal processing unit.

Claims

1. Sensor (10) for detecting at least one property of a fluid medium (12) in at least one measurement space (14), in particular for detecting an H2 fraction in a measurement gas (16), comprising at least: a first sensor element (18), which is designed to detect a thermal conductivity of the fluid medium and to output a first measurement signal, a second sensor element (20), which comprises a semiconducting metal oxide and is designed to output a second measurement signal, a third sensor element (22) for detecting a physical property of the fluid medium, the third sensor element (22) differing from the first sensor element (18) and the second sensor element (20) in terms of the detected physical property and being designed to output a third measurement signal, and an electronic evaluation unit (28) for evaluating the first measurement signal, the second measurement signal and the third measurement signal, the electronic evaluation unit (28) furthermore being designed to output control signals to the first sensor element (18), the second sensor element (20) and the third sensor element (22) and to change operating parameters of the first sensor element, the second sensor element (20) and / or the third sensor element (22) by means of an algorithm, characterized in that the algorithm (30) comprises an artificial neural network, the electronic evaluation unit (28) being designed to output correction signals to the first sensor element (18), the second sensor element (20) and the third sensor element (22) to change the operating parameters.

2. Sensor (10) according to one of the preceding claims, the third sensor element (22) being designed to detect at least one physical property selected from the group consisting of: moisture, in particular relative humidity, pressure, in particular air pressure, and temperature, in particular air temperature.

3. Sensor (10) according to either of the preceding claims, the first sensor element (18), the second sensor element (20) or the third sensor element (22) being separate sensor elements.

4. Sensor (10) according to one of Claims 1 to 3, the first sensor element (18), the second sensor element (20) or the third sensor element (22) being sensor elements integrated in a sensor chip.

5. Sensor (10) according to one of the preceding claims, furthermore comprising a voltage converter (32), the voltage converter (32) being designed for connection to an external voltage source, the voltage converter (32) furthermore being designed to generate a supply voltage for the first sensor element (18), the second sensor element (20) and the third sensor element (22).

6. Sensor (10) according to one of the preceding claims, furthermore comprising an interface (34), the interface (34) being designed to receive control commands from an external control unit and / or to output measurement data of the sensor to an external control unit.

7. Sensor (10) according to one of the preceding claims, furthermore comprising a sensor housing (36), the first sensor element (18), the second sensor element (20), the third sensor element (22) and the electronic evaluation unit (28) being arranged in the sensor housing (36), the sensor housing (36) having at least one opening (38) by means of which the first sensor element (18), the second sensor element (20) and the third sensor element (22) can be exposed to the fluid medium.