Gas detection device comprising plurality of detectors for different target gases

Abstract

The present invention relates to a gas detection device (100) and a gas detection method capable of detecting N types of target gases in a gas sample. A gas sample is supplied into the measurement chamber. The M detectors each generate a signal relating to the concentration of at least one of the N target gases (CH4, C3H8, C4H10) to be detected in the gas sample (Gp). The determiner (9) comprises M inputs (E.1, E.2, E.3) and N outputs (A.1, A.2, A.3). A signal (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6) is applied at each input (E.1, E.2, E.3) of the determiner (9), which signal depends on the signal of the associated measurement detector (5.1, 5.2, 5.3). Each output (A.1, A.2, A.3) supplies information about the concentration of the associated target gas (CH4, C3H8, C4H10). A determiner (9) is trained by applying a learning method to a sample having a plurality of sample elements. Each sample element contains M values for M signals from the measurement detector and N values for N target gas concentrations.

Description

Technical Field

[0001] This invention relates to a gas detection device comprising multiple measuring detectors and capable of detecting various target gases in a gas sample. Furthermore, this invention includes a gas detection method performed using such a gas detection device. Background Technology

[0002] Typical applications of gas detection devices include: at least one target gas (optionally, multiple target gases simultaneously) may be present in a space. In one application, the target gas, or each target gas, is hazardous (harmful) to humans, for example, because it is flammable or toxic. At least one target gas may also be a gas essential to life, particularly oxygen or an anesthetic. The gas detection device is capable of detecting target gases or each target gas in a given group of target gases and optionally determining the relevant concentration of each target gas. Summary of the Invention

[0003] The present invention aims to provide a gas detection device and a gas detection method that can simultaneously detect multiple predetermined target gases in a gas sample and is more reliable than known gas detection devices and methods in many cases.

[0004] The objective is achieved by the gas detection device and the gas detection method of the present invention. Advantageous embodiments are specified in this invention. Advantageous embodiments of the gas detection device according to the invention are also advantageous embodiments of the gas detection method according to the invention, where applicable, and vice versa.

[0005] A target gas group is defined as having N different target gases. Therefore, it is determined which N different target gases will be detected. The task of detecting the target gases can lead to the current exclusion of the presence of a target gas. N is a predetermined number greater than or equal to 2. The gas detection device and the gas detection method according to the invention are capable of detecting each of these N target gases in a gas sample, and preferably are also capable of determining the corresponding concentration of each target gas to be detected.

[0006] The gas detection device according to the invention includes a measuring chamber. This measuring chamber is capable of containing (holding) a gas sample to be inspected. The gas sample may contain no target gas, or it may contain at least one of N target gases. Preferably, the gas sample originates from the space region to be monitored.

[0007] Furthermore, the gas detection device includes a detection arrangement. The detection arrangement includes M measuring detectors, where M is a predetermined number greater than or equal to 2 and preferably greater than or equal to 3. The number M can be greater than, equal to, or less than N, and is preferably at least as large as N. Each measuring detector in the detection arrangement is capable of generating a corresponding signal. This signal is related to (depending on) the concentration of at least one of N target gases in a gas sample contained in a measuring chamber. Therefore, the signal is influenced by the concentration of at least one target gas, preferably by the concentration of each target gas. Conversely, the gas detection device is configured such that each target gas to be detected influences the signal of at least one measuring detector in the detection arrangement. The signal of this measuring detector depends on the concentration of that target gas in the gas sample. Generally, the signal does not depend on the concentration of other target gases or at least one other target gas in the gas sample, and is therefore not influenced by those other target gases.

[0008] The detection setup is configured to achieve the following: - If the gas sample does not contain any of the target gas to be detected, each measurement detector generates a reference signal. "Does not contain the target gas" means that the concentration of the target gas is below the construction-related detection limit. However, if the gas sample has different chemical compositions, at least one measuring detector will generate a signal that deviates from its reference signal. "Different chemical compositions" means that at least one of the N target gases is present in the gas sample at a concentration above the detection limit. The gas sample then contains at least one target gas to be detected.

[0009] The gas detection device also includes a signal processing determiner (determining unit). Generally, the determiner is implemented as software and can run on a processor. It can also be a signal processing unit. For each measuring detector in the detection arrangement, the determiner includes a corresponding input, thus there are a total of M different inputs (M is the number of detectors). Therefore, each input is associated with a measuring detector. Optionally, the determiner includes at least one additional input. Furthermore, the determiner includes an output for each target gas to be detected, i.e., a total of N different outputs for specifying the N target gases in the target gas group.

[0010] Gas detection methods utilize this type of gas detection device.

[0011] The gas detection device is configured as follows, and the gas detection method includes the following steps: a signal generated by or produced by a measuring detector is applied to an input of a determiner associated with the measuring detector. More generally: a signal is applied to this input, wherein the applied signal depends on the signal of the associated measuring detector. A signal from a reference detector or a sensor for environmental conditions is present at a corresponding optional additional input.

[0012] The output of the determiner associated with the target gas in the target gas group provides a determination value that depends on the concentration of the target gas in the gas sample. This value may be, for example, the target gas concentration itself, or a suitably normalized or standardized target gas concentration, or an indication of whether the target gas concentration is above a predetermined lower concentration limit (i.e., whether it exists). In one embodiment, the signal from the measuring detector is normalized or standardized using the signal from an optional reference detector described below.

[0013] Therefore, the determiner receives M signals from M measuring detectors at its M inputs, and optionally receives another signal at at least one additional input, such as a signal from an optional reference detector and / or a sensor for environmental conditions. The determiner provides N values ​​for N concentrations at its N outputs, where the N target gases to be detected have these concentrations in the gas sample. Of course, the gas sample may not contain at least one target gas to be detected, or even any of the N target gases.

[0014] The determiner is created as follows: a learning method is applied to the training samples. This application automatically trains the determiner. The training samples used for training consist of multiple sample elements. The corresponding chemical composition of each sample element and the M signals from the M measurement detectors for that chemical composition are known. Each sample element includes M values ​​on one hand, i.e., a corresponding value for the M measurement detectors, and on the other hand, N values, i.e., a corresponding value for the N target gases. The value for the measurement detector depends on the signal generated by that measurement detector. The value for the target gas depends on the concentration of that target gas in the gas sample. Each sample element refers to a specific combination of the N target gas concentrations. If the gas sample has this combination of the N target gas concentrations, then each measurement detector generates that signal.

[0015] Training samples are generated in advance. During generation, corresponding gas samples with known chemical compositions are used for each sample element. "Known chemical composition" means that the relevant concentration of each target gas in each gas sample is known. M signals from M measuring detectors are generated using the same gas detection device subsequently used to analyze unknown gas samples, or using a gas detection device with the same design and implementation.

[0016] The determinant is created in advance during the training phase. It is possible to retrain the determinant after using the gas detection device. Specifically, through retraining, the determinant adapts to changes in the state or operating conditions of the gas detection device.

[0017] The gas detection method according to the present invention includes corresponding steps.

[0018] Generally, for specific applications, it is known which target gases are present in the space to be monitored and therefore in the gas sample from that space. Thus, the N target gases are usually known in advance. Due to this invention, it is not necessary to use at least two different gas detection devices to detect the N target gases. Instead, the gas detection device according to the invention can simultaneously examine the gas sample in the measuring chamber for the N different target gases and provide the relevant target gas concentration for each. The gas detection device can automatically determine, for at least each of the N target gases, whether the gas sample in the measuring chamber contains that target gas at a concentration above a predetermined lower concentration limit (particularly the detection limit).

[0019] The gas detection device according to the invention comprises M measuring detectors. The determiner includes a corresponding input for each measuring detector and a corresponding output for each target gas to be detected. In particular, due to this feature, all measuring detectors can potentially use the same measurement principle. This advantage facilitates the manufacture and maintenance of the gas detection device compared to gas detection devices comprising at least two measuring detectors with different operations.

[0020] Furthermore, in many cases, it is possible to use at least one component of the gas detection device for all M measuring detectors. Generally, a single measuring chamber is sufficient to collect gas samples and analyze them for N target gases. In the case of photoelectric (photonic) or photoacoustic measuring detectors, it is possible to use a single radiation or sound source for all measuring detectors. This would be impossible if a measuring detector customized for each target gas and, for example, applying a measurement principle particularly suitable for that target gas had to be used. The gas detection device according to the invention requires only one display unit and only one single connection to the static power network or only one self-powered unit.

[0021] Because of this invention, it is unnecessary to switch the gas detection device between different modes during operation, each associated with a target gas, and wherein the gas detection device is configured to detect the target gas when operating in the mode associated with the target gas, and optionally configured to measure the concentration of the target gas. If the user sets these modes, the setting will require user action. If the user forgets the mode or cannot set the mode due to defects, there is a risk of missing harmful target gases or target gases essential to life. If the gas detection device will automatically switch from one mode to another, this will require the corresponding control unit, which may be malfunctioning or faulty. In both embodiments, procedures with different modes require more time to analyze gas samples for each target gas to be detected compared to the time required by the gas detection device according to the invention. Therefore, the gas detection device according to the invention has a shorter reaction time. In addition, readjustment may sometimes be necessary.

[0022] According to the invention, the determiner is generated using training samples. Therefore, in many cases, it is unnecessary to pre-analyze how the target gas affects the detection setup. Instead, the training samples are generated using the same gas detection device or at least a gas detection device with the same design and / or implementation as the one later used to analyze gas samples in the measurement chamber. Furthermore, in many cases, the results from the gas detection device according to the invention depend relatively little on the cross-sensitivity between different target gases.

[0023] As already explained, the determiner is trained using training samples. The training samples comprise multiple sample elements. Each sample element comprises a gas sample with a known chemical composition and includes a corresponding signal from each measurement detector. In one embodiment, the correlated gas sample from the multiple sample elements (preferably from each sample element) contains only one target gas with a concentration above a lower concentration limit (e.g., a detection limit). The correlated signal provided by the measurement detector for that sample element is divided by the concentration of the target gas contained in the gas sample of that sample element. The sample element also includes a vector of values ​​(i.e., a corresponding value for the concentration of the correlated target gas). In the embodiment just described, this vector consists of 1s and N-1 zeros. Optionally, the correlated signal from the measurement detector is additionally divided by the signal from an optional reference detector or otherwise normalized using the reference detector signal. The reference detector provides this signal for that gas sample.

[0024] Different embodiments are possible regarding how the determiner is generated using training samples. In one embodiment, the determiner comprises a neural network. This neural network is generated using training samples. Methods known in the art for training neural networks can be applied to this invention.

[0025] However, in preferred embodiments, the fact that in many cases, the particular effect of each target gas on the signal of the measuring detector does not depend in a correlated manner on the effect of another target gas (relatively low cross-sensitivity) is utilized. With sufficient accuracy, it is often assumed that the absorption spectra of the target gases are linearly superimposed. Therefore, a certain combination of signal values ​​from M measuring detectors is often caused (influenced) solely by a mixture of N target gases with a certain mixing ratio. This also applies if the mixture does not contain at least one of the N target gases at all. Thus, the combination of signal values ​​is a “fingerprint” of the gas mixture with that mixing ratio. Also with sufficient accuracy in many cases, the effect of the target gas on the measuring detector can be described as a weighted combination of the target gas concentrations in the gas sample, where the weighting factor of this weighted combination depends on the absorption behavior of the target gases and the design and implementation of the measuring detector, and is therefore constant in use.

[0026] According to this embodiment, training samples are used, and a matrix is ​​generated during the training of the determiner. A This makes the product, for the training samples, A x with vector b The deviation should be as small as possible, ideally it should be the same as b Match. Here, A It is a matrix consisting of M rows and N columns; that is, one row represents the M measuring detectors, and one column represents the N target gases to be detected. Vector x It consists of N vector elements, each depending on the corresponding concentration of one of the N target gases. b It includes M vector elements, which depend on the M signals from the M measurement detectors. Preferably, the vector... b Each element depends exactly on a corresponding measurement detector. Vectors are obtained in advance using training samples. x and b And based on vectors x and b Generating matrix A .matrix A It is generated before the gas detection device is used and stored, for example, in the memory of the gas detection device.

[0027] When using a gas detection device, a vector is generated based on M signals from M measuring detectors. b And using equations Generate a vector for seeking the concentration of N target gases to be detected. x Here, A -1 It is a matrix A The inverse matrix or pseudo-inverse matrix.

[0028] As is well known, neural networks are black boxes. It is often difficult to understand why such black boxes deliver a certain result. In particular, implementations with matrices have the following advantages compared to neural networks: - Compared to the weighting factors and activation functions of neural networks, matrices elements and vectors b and x They all have physical units of measurement and physical meanings. -Inspectable matrix The rationality of it. In many cases, it is possible to infer how and why the determiner and therefore the gas detection device provide a certain result regarding the concentration of the target gas. This derivation can be output in at least one form that is perceptible to humans (at least for the verification of the device) by means of a suitable output unit.

[0029] The signals from the M measuring detectors are often influenced not only by the chemical composition of the gas sample in the measuring chamber, but also by environmental conditions (particularly ambient temperature, ambient humidity, and / or ambient pressure), and by the current state of the components of the gas detection device (e.g., the state of the radiation source or optical filter or its own power supply unit). Therefore, the gas detection device preferably additionally includes a reference detector. The reference detector may also generate a signal. This signal is influenced by the ambient conditions and / or the current state of the gas detection device, but ideally, is independent of the chemical composition of the gas sample in the measuring chamber. Ideally, the M measuring detectors respond to the current state and ambient conditions in the same manner as the reference detector. The signal from the reference detector can be used to computationally compensate (at least approximately compensate) for the effects of the environmental conditions and the current state. Embodiments with a reference detector eliminate the need for direct measurement of environmental conditions. However, it is also possible for the gas detection device to include a sensor for ambient conditions (particularly for ambient temperature), or to be configured to receive and process signals from a spatially distant sensor for ambient conditions. In one implementation, the determiner includes an additional input for the measured ambient conditions.

[0030] According to the invention, the M inputs of the determiner are associated with (assigned to) M measurement detectors. In one embodiment, the signal of a particular measurement detector is directly present at the input associated with that measurement detector, wherein the present signal depends on the signal of that measurement detector, for example, directly on the signal of the measurement detector. In another embodiment, the M signals present at the M inputs of the determiner additionally depend on the signal of a reference detector. In a first alternative, the larger the signal of the associated measurement detector, the larger the signal present at the input, while the signal of the reference detector remains constant; and the larger the signal of the reference detector, the smaller the signal present at the input, while the signal of the measurement detector remains constant. In a second alternative, conversely, the larger the signal of the associated measurement detector, the smaller the applied signal, and the larger the signal of the reference detector, the larger the applied signal. For example, the quotient of the signal of the measurement detector (numerator) and the signal of the reference detector (denominator) exists at the associated input.

[0031] In this embodiment, M signals are applied to the determiner, and ideally, the effects of environmental conditions and the current state of the gas detection device have been computationally compensated for. This embodiment often leads to better results compared to cases where there are unadjusted signals from the measuring detector at the input.

[0032] According to the invention, the determiner includes M inputs for M measurement detectors. In one embodiment, the determiner includes additional inputs, namely, inputs regarding a reference detector. This embodiment makes it possible to compensate for the effects of environmental conditions and / or the current state that have not been eliminated by the embodiments described above (i.e., particularly not by training).

[0033] Alternatively, the gas detection device receives signals for environmental conditions from a corresponding sensor, and the determiner has another input for the signal from that sensor. For example, the neural network is additionally trained using signals from a reference detector and / or signals from the sensor for environmental conditions. Optionally, the determiner also includes an input of the signal from the sensor for environmental conditions.

[0034] According to the present invention, a gas detection device includes M measuring detectors and, optionally, a reference detector. In one embodiment, the M measuring detectors and the optional reference detector are all implemented as photoelectric (photonic) or photoacoustic receivers. The gas detection device includes a radiation source. The radiation source emits electromagnetic radiation or sound, such as ultrasound. Hereinafter, the abbreviation "electromagnetic radiation" is used, and this term also includes sound and ultrasound. The same radiation source and the same voltage source used for the radiation source are used for all measuring detectors and the optional reference detector. Preferably, the gas measuring device includes a measuring chamber and a reference chamber that does not contain each target gas.

[0035] According to an embodiment with a receiver, the gas detection device is configured such that at least a portion of the emitted electromagnetic radiation penetrates the measurement chamber, and thus penetrates the gas sample in the measurement chamber at least once and collides with the measurement detector. Optionally, the electromagnetic radiation is reflected at least once before colliding with the measurement detector to extend the optical path. In one embodiment, another portion of the electromagnetic radiation is guided through the measurement chamber, and thus through the gas sample, preferably through a reference chamber, and collides with a reference detector. In different embodiments, a wavelength filter upstream of the reference detector only allows radiation at frequencies at which the target gas does not attenuate the radiation to pass through. Each detector generates a signal that depends on the intensity of the incident electromagnetic radiation.

[0036] Each target gas to be detected causes an attenuation of electromagnetic radiation penetrating a gas sample containing that target gas within its target gas frequency band, where "attenuation" refers to the degree of attenuation above a lower attenuation limit. The degree of attenuation depends on the target gas concentration, while the target gas frequency band typically depends on the type of target gas, but to a relatively small extent on the target gas concentration, or even not at all. This target gas frequency band is often characteristic of the target gas and is typically known for each target gas to be detected, and therefore also known for N specified target gases.

[0037] The embodiments described above and below utilize this fact. A total frequency band is defined. The relevant target gas frequency band of the target gas to be detected lies within this entire frequency band. The radiation source is configured such that the electromagnetic radiation emitted by the radiation source covers the entire frequency band. This means that at any frequency at which at least one target gas to be detected causes radiation attenuation above a lower attenuation limit, the radiation source emits radiation with an intensity higher than a lower intensity limit. It is possible (but not necessarily) that the radiation has an intensity greater than a lower intensity limit at every frequency within the total frequency band.

[0038] According to the invention, the detection arrangement is configured such that a target gas in a gas sample causes at least one measuring detector to generate a signal deviating from a reference signal of that measuring detector. This applies to every target gas and every possible concentration of that target gas. For the embodiment just described, which has photoelectric or photoacoustic measuring detectors, this means that after radiation penetrates the gas sample at least once, the measuring detectors respond differently to the incident radiation. For example, in front of at least one measuring detector, preferably in front of each measuring detector, and additionally in front of an optional reference detector, there is a corresponding wavelength filter, wherein each wavelength filter allows only radiation within a specific frequency range to pass through (transmit). It is also possible that a prism or mirror divides the radiation between at least two measuring detectors, and in doing so, decomposes it into different frequency bands.

[0039] Preferably, all measurement detectors, and additionally, an optional reference detector, are constructed in the same manner and, as just described by example, generate at least two different signals. In one embodiment, different wavelength filters are arranged in front of the measurement detectors. This embodiment simplifies the fabrication of the measurement detectors. They can all be constructed in the same manner.

[0040] In one embodiment, the radiation source comprises at least two individual light sources, such as individual LEDs. Preferably, the number of individual light sources of the radiation source corresponds to the number M or greater of the measuring detectors in the detection arrangement. Each of the M inputs of the determiner assigns at least one individual light source of the radiation source.

[0041] Each individual light source is capable of emitting electromagnetic radiation within a specific light source frequency band. Within this frequency band, the intensity of the radiation emitted by that light source is above the lower intensity limit. The light source frequency bands together cover the total frequency band in the sense described above. This means that at any frequency at which at least one target gas causes radiation attenuation above the attenuation limit, the individual light sources together emit radiation with an intensity above the lower intensity limit. In many applications, this embodiment eliminates the need to emit radiation with an intensity greater than the lower intensity limit at every frequency across the entire frequency band. Therefore, this embodiment often saves electrical energy.

[0042] In embodiments where the radiation source comprises at least two individual light sources, the gas detection device is configured such that at any given time, at most one individual light source of the radiation source is turned on and emits electromagnetic radiation in its frequency band. The turned-on individual light source emits electromagnetic radiation continuously or in pulsed form. At this time, the other individual light sources of the radiation source, or any other individual light sources, are turned off. Preferably, each individual light source of the radiation source is temporarily turned on during the use of the gas detection device.

[0043] In many cases, this implementation eliminates the need for a wavelength filter between the radiation source and the measuring detector. Furthermore, this implementation eliminates the need for multiple individual measuring detectors spaced apart from each other. Instead, it is sufficient for the detection arrangement to include a single measuring detector. According to this training, only the individual light source is turned on at the moment the measuring detector supplies a signal value. Therefore, the signal value depends on how strongly the gas sample in the measuring chamber absorbs the electromagnetic radiation from this single, turned-on light source (i.e., radiation in the relevant light source frequency band).

[0044] Preferably, in the embodiment just described, each individual light source is associated with one of the M inputs of the determiner, and two different individual light sources are associated with two different inputs. According to the invention, the detection arrangement includes M measurement detectors. The following is recorded for the embodiment just described, wherein at any given time, at most one individual light source is on, and only one measurement detector is used: at any given moment when a particular individual light source is on, this measurement detector acts as the measurement detector associated with that particular one of the M inputs of the determiner. The signal provided by the measurement detector at this time acts as the signal of the measurement detector associated with that input.

[0045] The present invention further relates to an arrangement having a gas detection device and a signal processing generation device according to the invention, wherein the generation device is configured to automatically generate a determiner for the gas detection device according to the invention; and a method for generating the determiner.

[0046] The arrangement further includes a training gas sample set. The training gas sample set comprises multiple gas samples. The corresponding composition of each gas sample in the training gas sample set is known. This means that the corresponding concentration of each of the N target gases to be detected in the gas samples is known. For each target gas to be detected, the training gas sample set includes at least one gas sample containing that target gas at a concentration above a predetermined lower concentration limit. In a preferred embodiment, each target gas to be detected appears in at least 50, particularly preferably at least 100, different gas samples in the training gas sample set, these gas samples having concentrations above the lower concentration limit. It is possible that at least one gas sample in the training gas sample set (preferably, each gas sample) contains exactly one target gas at a concentration above the lower concentration limit, i.e., it does not contain any other target gases. Optionally, the training gas sample set additionally includes at least one gas sample that does not contain each target gas to be detected.

[0047] The generating apparatus is configured to perform the following steps for each gas sample in the training gas sample group, and the method for generating the determiner includes the following steps for each gas sample in the training gas sample group: - Supply the gas sample into the measuring chamber of the gas detection device. Preferably, the measuring chamber is pre-flushed. Each measurement detector generates a signal. This signal is related to the chemical composition of the gas sample (i.e., the N concentrations of the N target gases to be detected). Therefore, for a gas sample, there are M signals from the measurement detectors. - An optional reference detector also generates a signal. - Generate vector sample elements. On one hand, these vector sample elements include a corresponding identifier for the concentration of each target gas in the gas sample for each gas to be detected. As mentioned above, the chemical composition of the gas samples is known. On the other hand, these vector sample elements include a corresponding identifier for the signal generated by each measuring detector for that gas sample.

[0048] Furthermore, the generating apparatus is configured to perform the following steps, and the method includes the following steps:

[0049] The determiner is trained automatically by applying a learning method to the training samples. The training samples used for training consist of vector sample elements generated as just described (i.e., gas samples using the training gas sample set).

[0050] The advantages previously mentioned for this type of gas detection device also apply to this arrangement.

[0051] In one embodiment, the generating device is configured to generate a matrix. A , where the matrix A It includes M rows for M measuring detectors and N columns for N target gases. To generate the matrix... A The generator produces vector sample elements for each gas sample in the training gas samples, serving as the corresponding sample elements. Each vector sample element consists of a vector with N vector elements. x and a vector with M elements b .vector b The M vector elements depend on the signals from the M measurement detectors provided for the gas sample, and the vector... x The N vector elements depend on the concentrations of the N target gases in the gas sample. The generating device generates a matrix such that, for each element of the vector sample, the product... with vector b The deviation should be minimized. To this end, the generator uses vector sample elements. To generate the matrix... A The generating device preferably employs an optimization method that applies an error function, more preferably an iterative optimization method. The error function generates a product for each vector sample element. with vector b An indicator of the difference between them. Attached Figure Description

[0052] The invention is described below based on exemplary embodiments. In the accompanying drawings, Figure 1 A first embodiment of the gas detection device according to the invention is schematically shown, wherein a radiation source emits broadband electromagnetic radiation, and wherein the radiation strikes three parallel measuring detectors; Figure 2A second embodiment of the gas detection device according to the invention is schematically shown, wherein a plurality of diodes emit narrowband electromagnetic radiation, and as in Figure 1 In the middle, radiation impacts three parallel measurement detectors; Figure 3 A third embodiment of the gas detection device according to the invention is schematically shown, wherein, as in Figure 2 In this process, multiple diodes emit narrowband electromagnetic radiation, which then strikes a single measurement detector. Figure 4 A signal processing determinist with a neural network; Figure 5 For training the determinant, a generation matrix is ​​used; Figure 6 According to Figure 5 The determinant is trained using the pseudo-inverse of the matrix. Detailed Implementation

[0053] Figures 1 to 3 Two different embodiments of the gas detection device 100 according to the present invention are schematically illustrated. In all three embodiments, the gas detection device 100 according to the present invention is capable of simultaneously detecting N target gases. N is a predetermined number greater than or equal to 2, and is equal to 3 in the exemplary embodiment.

[0054] The gas sample Gp from the monitored space region is located in measurement chamber 2. Each target gas attenuates electromagnetic radiation eS in its corresponding target gas frequency band. The target gas frequency band may include at least two separate and independent frequency bands. The entire frequency band is defined. The relevant target gas frequency bands in which the target gas attenuates electromagnetic radiation eS are located are within the same entire frequency band.

[0055] Radiation source 1 emits electromagnetic radiation eS. The electromagnetic radiation eS emitted by radiation source 1 covers the entire frequency band. More specifically, for each frequency in the entire frequency band, this means the following: if at least the sum of the absorption coefficients of all N target gases to be detected at that frequency is greater than a predetermined lower absorption limit, then radiation source 1 emits electromagnetic radiation eS with a radiation intensity (strength) above the predetermined lower absorption limit at that frequency. This lower absorption limit is selected such that absorption above the lower absorption limit can be reliably detected.

[0056] A portion of the emitted electromagnetic radiation eS enters measurement chamber 2 and penetrates the gas sample Gp in measurement chamber 2 at least once. Optionally, the electromagnetic radiation eS is reflected at least once and penetrates the gas sample Gp in measurement chamber 2 at least twice. After penetrating the gas sample Gp at least once, the electromagnetic radiation eS impacts (collides with) three measurement detectors 5.1, 5.2, and 5.3 arranged parallel to the radiation eS. Another portion of the electromagnetic radiation eS passes through measurement chamber 2, and thus through the gas sample Gp and through reference chamber 3, and impacts reference detector 6.

[0057] Three measuring detectors 5.1, 5.2, 5.3 and reference detector 6 belong to detector arrangement 10.

[0058] According to Figure 1 and Figure 2 In this embodiment, each measurement detector 5.1, 5.2, 5.3 is excited by incident electromagnetic radiation eS only in a specific detector band. In other words, the signal of measurement detectors 5.1, 5.2, 5.3 is greater than a predetermined lower limit only when the frequency of the incident radiation is in the relevant detector band. For example, this characteristic is achieved by an appropriate wavelength filter (not shown) upstream of the relevant measurement detectors 5.1, 5.2, 5.3. The detector band also covers the entire frequency band. Apart from the wavelength filter, the three measurement detectors 5.1, 5.2, 5.3 can be constructed in the same manner.

[0059] Three measuring detectors 5.1, 5.2, and 5.3 each generate corresponding signals Sg.1, Sg.2, and Sg.3; reference detector 6 generates signal Sg.6. Each signal Sg.1, Sg.2, Sg.3, and Sg.6 is related to the intensity of the incident (impact) electromagnetic radiation eS.

[0060] According to Figure 1 In the first embodiment, the radiation source 4 (e.g., an IR diode) emits broadband electromagnetic radiation eS within the radio frequency band Fb. Figure 1 Figure S.1.1 schematically and ideally shows the spectrum S(f) of the broadband radiation eS covering the entire frequency band.

[0061] According to Figure 2 In the second embodiment, the three diodes 4.1, 4.2, and 4.3 arranged in parallel all emit corresponding electromagnetic radiation eS within the corresponding spectrum S(f). Figure 2Figure S.1.2 shows the superposition of these three spectra. The superposition also covers the entire frequency band. The advantages of the second embodiment are as follows: In many cases, the second embodiment requires less electrical energy than the first embodiment. One reason for this is that the emitted electromagnetic radiation eS does not necessarily have to cover the entire frequency band, but only those ranges within the entire frequency band where at least one of the target gases being detected absorbs radiation sufficiently strongly. In many cases, the second embodiment can meet this requirement using less electrical energy than the first embodiment. This advantage is particularly important if the gas detection device 100 is not permanently connected to a static power supply network, but instead includes its own power supply unit.

[0062] Of course, radiation sources other than infrared light sources are also possible, such as ultraviolet light sources or ultrasonic sources.

[0063] Figure S.2 illustrates, by way of example, the intensity of electromagnetic radiation eS if broadband radiation eS is emitted as shown in Figure S.1.1 and after the electromagnetic radiation eS penetrates the gas sample Gp. Figures S.3.1 and S.3.2 show the intensity of electromagnetic radiation eS incident (impacted) on measurement detectors 5.1, 5.2, and 5.3. The corresponding output signals Sg.1, Sg.2, and Sg.3 of measurement detectors 5.1, 5.2, and 5.3 depend on the corresponding concentration of each target gas. As already mentioned, the wavelength filter ensures that measurement detectors 5.1, 5.2, and 5.3 are only slightly excited or even not excited at all by incident electromagnetic radiation eS outside the detector band.

[0064] Reference detector 6 also generates a signal Sg.6, which depends on the intensity of the incident electromagnetic radiation eS. Ideally, the signal Sg.6 of reference detector 6 is completely independent of the concentration of the target gas in the monitored space region. Ideally, the signal Sg.6 of reference detector 6 is affected by environmental influences (such as ambient pressure, ambient temperature, and ambient humidity) and the current state of gas detection device 100 (e.g., radiation source 1 and optional power supply unit) in the same way as the corresponding signals of measuring detectors 5.1, 5.2, and 5.3. Therefore, the effect of environmental influences and the state of gas detection device 100 on the corresponding signals of measuring detectors 5.1, 5.2, and 5.3 is calculated using the signal from reference detector 6. For example, quotients Sg.1 / Sg.6, Sg.2 / Sg.6, and Sg.3 / Sg.6 are calculated and used. This quotient is referred to as the normalized signal of the corresponding measuring detectors 5.1, 5.2, and 5.3.

[0065] According to Figure 1 and Figure 2In both embodiments, the detector arrangement 10 includes three measurement detectors 5.1, 5.2, and 5.3 arranged in parallel. Each measurement detector 5.1, 5.2, and 5.3 is excited by incident electromagnetic radiation eS only within a specific detector frequency band. According to... Figure 1 In the first embodiment, radiation source 4 emits broadband electromagnetic radiation eS. According to... Figure 2 In the second embodiment, three parallel radiation sources 4.1, 4.2, and 4.3 emit narrowband electromagnetic radiation eS in three different spectra. The three radiation sources 4.1, 4.2, and 4.3 may emit electromagnetic radiation eS continuously or in a pulsed and time-overlapping manner.

[0066] Figure 3 A third alternative embodiment is shown. As in [the previous section]... Figure 2 Similar to the embodiments described above, the gas detection device 100 in the third embodiment also includes three parallel radiation sources 4.1, 4.2, and 4.3, which emit electromagnetic radiation eS in three different spectra. In contrast, the detector arrangement 10 includes a single measurement detector 5 and a reference detector 6, instead of three parallel measurement detectors 5.1, 5.2, and 5.3. This single measurement detector 5 also generates an output signal depending on the intensity of the incident electromagnetic radiation eS. A wavelength filter is not necessarily positioned between the radiation sources 4.1, 4.2, and 4.3 and the measurement detector 5.

[0067] According to the third embodiment, the three radiation sources 4.1, 4.2, and 4.3 can be turned on and off independently of each other. The turned-on radiation sources 4.1, 4.2, and 4.3 emit continuous or pulsed electromagnetic radiation. At any given time, no more than one radiation source 4.1, 4.2, or 4.3 is turned on while the other radiation sources are turned off. When the gas detection device 100 monitors the space area, each of the three radiation sources 4.1, 4.2, and 4.3 is temporarily turned on.

[0068] To use consistent notation across all three embodiments, the following notation is used for the third embodiment: Measurement detector 5 is designated as 5.x when radiation source 4.x is on and each of the other radiation sources 4.y is off. Sg.x represents the signal supplied by measurement detector 5 if radiation source 4.x is on and when radiation source 4.x is on (x = 1, 2, 3, y ≤ x).

[0069] The gas detection device 100 includes a signal processing and evaluation unit 8, see [link to relevant documentation]. Figure 4 For each target gas to be detected, the evaluation unit 8 provides at least a statement (information) regarding whether the target gas is present in gas sample Gp at a concentration above a predetermined lower concentration limit (optionally, the corresponding concentration of the target gas in gas sample Gp). The target gas concentration determined by the evaluation unit 8 is typically only an approximation of the actual (true) target gas concentration in the space region to be monitored.

[0070] The normalizer 7 of the signal processing evaluation unit 8 normalizes the three signals from the three measurement detectors 5.1, 5.2, and 5.3. For example, the normalizer 7 divides the corresponding signals of measurement detectors 5.1, 5.2, and 5.3 by the signal of the reference detector 6. The determiner 9 of the evaluation unit 8 includes one input for each measurement detector 5.1, 5.2, and 5.3, and optionally an additional input for the reference detector 6.

[0071] According to Figures 4 to 6 In both embodiments, the determiner 9 has four inputs E.1, E.2, E.3, E.6. The corresponding normalized signals Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6 of the measurement detectors 5.1, 5.2, 5.3 are located at one of the first three inputs E.1, E.2, E.3, while the signal Sg.6 of the reference detector 6 is located at the fourth input E.6. The input E.6 for the signal Sg.6 of the reference detector 6 is optional.

[0072] In the example shown, the gas detection device 100 is configured to detect three combustible target gases in the gas sample Gp: methane (CH4), propane (C3H8), and butane (C4H4). 10 In the example shown, the determiner 9 has three outputs A.1, A.2, and A.3, for each target gas to be detected: CH4, C3H8, and C4H. 10 One output is A.1, A.2, A.3. The target gas, the number of inputs, and the number of outputs are merely examples. Of course, other target gases and different numbers of outputs are also possible. Determiner 9 provides the determined concentration of the associated target gas at the correspondingly assigned outputs A.1, A.2, A.3.

[0073] Based on Figure 4 In the embodiments, further according to Figure 5 and 6In this embodiment, the determiner 9 is pre-trained using computer-evaluable training samples. These training samples comprise multiple sample elements. Each sample element is generated using a gas sample Gp. The gas sample Gp used to generate the sample element has a known chemical composition. This means that the corresponding concentration of each target gas to be detected in the gas sample Gp is ​​known. Each gas sample Gp is ​​supplied or otherwise delivered to the measurement chamber 2, where the radiation source 1 emits electromagnetic radiation eS, and the result signals Sg.1, Sg.2, Sg.3, Sg.6 for each detector 5.1, 5.2, 5.3, 6 are recorded and stored. Preferably, for each target gas to be detected, at least K gas samples are generated and measured, where K is greater than or equal to 100, and these K gas samples contain only that one target gas and do not contain the target gas or any other target gas to be detected. However, it is also possible that the gas sample of at least one sample element contains two different target gases with sufficiently high target gas concentrations.

[0074] N represents the number of target gases to be detected, and M represents the number of measuring detectors. In this case, M = N = 3 holds true. Each sample element includes N values ​​for the concentrations of the N target gases in a gas sample Gp, which has a known chemical composition that results in that sample element. Furthermore, each sample element includes M values ​​for the M signals Sg.1, Sg.2, Sg.3 generated by the M measuring detectors when analyzing the gas sample. By way of example, Figures 4 to 6 The three normalized signal values ​​Sg.1 = 0.8, Sg.2 = 0.8, and Sg.3 = 0.2 for the three measurement detectors 5.1, 5.2, and 5.3, and the signal value Sg.6 = 0.9 for the reference detector 6 are shown.

[0075] exist Figures 4 to 6 In the example shown, each gas sample Gp used to generate training samples has only one target gas with a concentration above the lower limit. In the example shown, on one hand, each signal Sg.1, Sg.2, Sg.3 of measuring detectors 5.1, 5.2, 5.3 is normalized using the signal of reference detector 6. On the other hand, the normalized signals Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6 are divided by the relevant target gas concentration. Therefore, the N values ​​for the target gas concentration are equal to 1 once and equal to 0(N-1) times. Through the example, Figure 4 and Figure 5 The value 1 is shown for methane, and the value 0 is shown for butane and propane.

[0076] exist Figure 4In this example, classifier 9 comprises a neural network. This neural network is trained using training samples. The neural network has M inputs E.1, E.2, E.3 for M measurement detectors 5.1, ..., an optional additional input E.6 for reference detector 6, and N outputs A.1, A.2, A.3 for N target gases to be detected. Neural network 9 includes an input layer Li, an intermediate layer Lm, and an output layer Lo. The neural network can be trained using a backpropagation algorithm. Using backpropagation is preferred. Figure 4 An example of a node (functional unit, "perceptron") of a neural network is shown in the lower right corner. The node has n inputs and forms a weighted average of the n values ​​at its n inputs using n weighting factors w1,...,wn. The node applies a non-linear function (shown via the example) to this weighted average. For example, this function is a thresholding function, or more generally, an activation function. Optionally, the result is fed back once using a weighting factor wf.

[0077] Preferably, the training samples are divided into two groups of sample elements. The first group is used to train the neural network 9. The second group is used to check the trained neural network 9. If the deviation between the result of the neural network 9 and the actual target gas concentration is too large, then retraining is preferably performed using another (or even more) training samples.

[0078] The advantages of implementing classifier 9 using a neural network are as follows: There is no need to make assumptions about the effects of different target gases on the gas sample Gp. Furthermore, there is no need to make assumptions about the cross-sensitivity or lack thereof of the signals from the measurement detectors 5.1, 5.2, and 5.3 regarding different target gases. One disadvantage is that neural network 9 is essentially a black box. The weighting factors and activation functions generally do not have physical meaning. It is impossible to interpret or analytically verify the results of neural network 9.

[0079] According to Figure 5 and Figure 6 In this embodiment, the following model assumptions are made: the target gases in the gas sample Gp attenuate the electromagnetic radiation eS of a certain frequency in the measurement chamber 9 independently of each other. The effects of the two different target gases on the electromagnetic radiation eS of a certain frequency are additive. There is no cross-sensitivity between the target gases. In many cases, these model assumptions are sufficiently accurate.

[0080] Figure 5 This demonstrates how to train the determinant 9. Figure 6 This demonstrates the application of the trained determinant 9. Based on... Figure 5 During training, a matrix is ​​generated by applying training samples. matrix This includes a corresponding row for each measurement detector 5.1, 5.2, 5.3, and for each target gas to be detected: CH4, C3H8, C4H 10 The corresponding column. Optionally, the matrix This includes an additional row for reference detector 6. In the example shown, the matrix... It has M+1 = 4 rows and N = 3 columns.

[0081] In this embodiment, sample elements are represented by vector sample elements. Each sample element is represented by a first vector having N vector elements. x (That is, one vector element for each target gas to be detected) and a second vector having M+1 vector elements for M measurement detectors 5.1, 5.2, 5.3 and for reference detector 6. b Composition. In the example shown, the signal value for the gas sample Gp is ​​input into the matrix. As described only, this value is normalized using the signal value Sg.6 of reference detector 6 and the target gas concentration using the single target gas in the gas sample (in this example: methane). Vector x It consists of 1 for methane and 0 for propane and butane in each case. Vector b It consists of three normalized signal values ​​Sg.1 / Sg.6, Sg.2 / Sg.6, and Sg.3 / Sg.6 from three measurement detectors 5.1, 5.2, and 5.3, and the signal value Sg.6 from reference detector 6.

[0082] In an alternative embodiment, at least one zero-point measurement and one zero-point calibration are performed beforehand. During the zero-point measurement, or during each zero-point measurement, a gas sample excluding the target gas is supplied to the measurement chamber 9. Also in this case, each measurement detector 5.1, 5.2, 5.3 is supplied with a signal value used as the zero value. These zero values ​​are determined by Sg.1. Null Sg.2 Null Sg.3 Null Sg.6 Null This indicates that the measurement provided is the raw signal value Sg1, describing only the measurements for different gas samples (each sample containing a single target gas). Roh Sg.2 Roh Sg.3 Roh Sg.6 Roh The difference Sg.i = Sg.i Roh -Sg.i Null (i = 1, 2, 3, 6) are used to determine the order of operations. Figures 4 to 6 The signal values ​​of the program.

[0083] Figure 5This illustrates the ideal scenario where each of the K gas samples containing only methane results in the same vector. b In fact, these K gas samples result in up to K distinct vectors. The same applies to any other target gas to be detected. Therefore, for each target gas i (i = 1,...,N), the average vector is calculated. b mean (i). To compute this vector b mean (i) Averages b using K vectors, which are generated as a measure of the reaction of K gas samples with target gas i. b mean (i) used as a matrix The i-th column represents the target gas i (i = 1, ..., N).

[0084] When using the gas detection device 100, a signal value vector is generated from the signals of M+1 detectors 5.1, 5.2, 5.3, and 6. b See Figure 6 The unknown and expected chemical composition of the gas sample Gp in measuring chamber 2 is determined by equation... Confirmed. Generally speaking, a matrix It is not symmetric. Matrix It is the pseudo-inverse of matrix A. Various methods can be used to compute the matrix. The pseudo-inverse matrix, especially Singular Value Decomposition (SVD) -Thikonov regularization -Kaczmarz algorithm.

[0085] In one embodiment, the pseudo-inverse matrix By analyzing the matrix The inverse is generated and saved once in advance, and used as part of determiner 9. In another embodiment, the pseudo-inverse matrix... The calculation is performed during runtime, preferably iteratively, until a predetermined termination criterion is met. List of reference numerals

Claims

1. A method for detecting N target gases (CH4, C3H8, C4H) in a gas sample (Gp). 10 Gas detection device (100), Among them, there are N target gases to be detected (CH4, C3H8, C4H). 10 ) is predetermined, and where N is a predetermined number greater than or equal to 2. The gas detection device (100) includes - Measurement chamber (2), - A detection arrangement (10) with M measurement detectors (5.1, 5.2, 5.3), and - An evaluation device (8) with a signal processing determinant (9), Where M is a predetermined number greater than or equal to 2. The measuring chamber (2) is configured to contain the gas sample (Gp) to be analyzed. Each of the measurement detectors (5.1, 5.2, 5.3) is configured to generate a signal (Sg.1, Sg.2, Sg.3), which is correlated with the N target gases (CH4, C3H8, C4H4) in the gas sample (Gp). 10 The concentration of at least one of them is related. The detection arrangement (10) is configured such that If the gas sample (Gp) does not contain any of the target gas to be detected, each measurement detector generates a reference signal, and For each different chemical component of the gas sample (Gp), at least one measuring detector (5.1, 5.2, 5.3) generates a signal that deviates from the reference signal of the measuring detector (5.1, 5.2, 5.3). The determiner (9) includes -M inputs (E.1, E.2, E.3), meaning one input for each measurement detector (5.1, 5.2, 5.3), and -N outputs (A.1, A.2, A.3), representing each target gas to be detected (CH4, C3H8, C4H). 10 An output (A.1, A.2, A.3) is given. The gas detection device (100) is configured such that - At each of the M inputs (E.1, E.2, E.3) of the determiner (9), a corresponding signal (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6) is applied, the signal depending on the corresponding signal (Sg.1, Sg.2, Sg.3) of the corresponding associated measurement detector (5.1, 5.2, 5.3), and - Each of the N outputs (A.1, A.2, A.3) provides information about the corresponding associated target gas to be detected (CH4, C3H8, C4H). 10 The information regarding the concentration of ) The determiner (9) is trained by applying a learning method to the training samples. The training samples used for this training include multiple sample elements. Each sample element includes -For each target gas to be detected (CH4, C3H8, C4H) 10 The value of the concentration of the target gas, and - The value of the signal (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6) generated by each measuring detector (5.1, 5.2, 5.3) for these N target gas concentrations.

2. The gas detection device (100) according to claim 1, Its features M is a predetermined number that is greater than or equal to N.

3. The gas detection device (100) according to claim 1, Its features The determiner (9) is configured to use equations The N target gas concentrations are provided at its N outputs (A.1, A.2, A.3). The vector b This includes M vector elements for M signals (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6), said signals depending on the M signals (Sg.1, Sg.2, Sg.3) of the M measurement detectors (5.1, 5.2, 5.3). The vector x Includes N vector elements that depend on the concentrations of the N target gases. The matrix It is a matrix The inverse matrix or pseudo-inverse matrix, and The matrix -Including M rows for the M measuring detectors (5.1, 5.2, 5.3) and N columns for the N target gases, and - Using the training samples and equations Ax = b generate.

4. The gas detection device (100) according to any one of the preceding claims, Its features The detection arrangement (10) includes a reference detector (6), The reference detector (6) is configured to generate a signal (Sg.6), the signal - Affected by ambient conditions and / or the state of the gas detection device (100), and -Irrelevant to the chemical composition of the gas sample (Gp) in the measuring chamber (2).

5. The gas detection device (100) according to claim 4, Its features The gas detection device (100) is configured such that The signals (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6) applied to the inputs (E.1, E.2, E.3) of the determiner (9) associated with the measurement detectors (5.1, 5.2, 5.3) depend on the signals (Sg.1, Sg.2, Sg.3) of the measurement detectors (5.1, 5.2, 5.3) and additionally on the signal (Sg.6) of the reference detector (6). in - In the first alternative, the larger the signal (Sg.1, Sg.2, Sg.3) of the measurement detector (5.1, 5.2, 5.3) and the smaller the signal (Sg.6) of the reference detector (6), the larger the signal (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6) applied to the input (E.1, E.2, E.3), and the larger the signal (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6). - In the second alternative, the smaller the signal (Sg.1,Sg.2,Sg.3) of the measurement detector (5.1,5.2,5.3) and the larger the signal (Sg.6) of the reference detector (6), the larger the signal (Sg.1 / Sg.6,Sg.2 / Sg.6,Sg.3 / Sg.6) applied to the input (E.1,E.2,E.3).

6. The gas detection device (100) according to claim 4, Its features The determiner (9) includes another input (E.6), namely the input for the reference detector (6). The gas detection device (100) is configured such that a signal (Sg.6) from the reference detector (6) is applied to the other input (E.6).

7. The gas detection device (100) according to any one of claims 1-3, Its features Each target gas to be detected (CH4, C3H8, C4H) 10 This causes the electromagnetic radiation (eS) in the corresponding target gas frequency band to attenuate. Define a total frequency band such that each target gas frequency band is included by the total frequency band, and The gas detection device (100) includes a radiation source (1), The radiation source (1) is configured to emit electromagnetic radiation (eS). The frequency band (Fb) of the emitted electromagnetic radiation (eS) covers the total frequency band. The gas detection device (100) is configured such that at least a portion of the emitted electromagnetic radiation (eS) penetrates the measuring chamber (2) at least once, and after penetration, impacts the detection arrangement (10). Each of the measurement detectors (5.1, 5.2, 5.3) is configured to generate a signal (Sg.1, Sg.2, Sg.3) that depends on the intensity of the incident electromagnetic radiation (eS).

8. The gas detection device (100) according to claim 7, Its features The radiation source (1) includes at least two separate light sources (4.1, 4.2, 4.3). Each individual light source (4.1, 4.2, 4.3) of the radiation source (1) is configured to emit electromagnetic radiation (eS) in the corresponding light source frequency band, and The light source frequency bands together cover the total frequency band in the following manner, such that For each frequency at which at least one target gas causes electromagnetic radiation (eS) to attenuate above a predetermined attenuation lower limit, The electromagnetic radiation (eS) emitted by the individual light sources (4.1, 4.2, 4.3) has a total intensity greater than a predetermined lower limit at this frequency.

9. The gas detection device (100) according to claim 8, Its features The gas detection device (100) is configured such that At any given time, at most one individual light source (4.1, 4.2, 4.3) of the radiation source (1) is turned on, and the radiation source (1) or any other individual light source is turned off.

10. An arrangement comprising -The gas detection device (100) according to any one of the preceding claims, - Generating device, and -A training gas sample set with multiple gas samples The corresponding chemical composition of each gas sample in the training gas sample group is known, and the target gas to be detected (CH4, C3H8, C4H) is also known. 10 ) appears in at least one gas sample, and The generating device is configured to perform the following steps for each gas sample in the training gas sample group: - To guide the gas sample into the measuring chamber (2) of the gas detection device (100), - Each measurement detector (5.1, 5.2, 5.3) generates a signal (Sg.1, Sg.2, Sg.3), which is compared with the N target gases (CH4, C3H8, C4H) in the gas sample (Gp). 10 The corresponding concentration of ) is related, and - Generate training sample elements, which include: For each target gas to be detected, the target gas (CH4, C3H8, C4H) in the gas sample (Gp) 10 The concentration indication of ) and The identifier for each measurement detector (5.1, 5.2, 5.3) of the signal (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6) generated by that measurement detector (5.1, 5.2, 5.3) for that gas sample (Gp). The generation device is further configured to train the determiner (9) by applying a learning method to training samples, and The training samples used for training include training sample elements generated using the gas samples from the training gas sample group.

11. The arrangement according to claim 10, Its features The generation device is configured to generate vector sample elements as corresponding training sample elements for each gas sample in the training gas sample group. The vector sample elements include vectors b sum vector x , The vector b This includes M vector elements for the M signals (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6), which depend on the M signals (Sg.1, Sg.2, Sg.3) generated by the M measurement detectors (5.1, 5.2, 5.3) for the gas sample. The vector x Includes N vector elements that depend on the concentrations of the N target gases in the gas sample. The generating device is further configured to generate a matrix. So that for the elements of the vector sample, the product x With the vector b The deviation should be as small as possible. The matrix It includes M rows for the M measuring detectors (5.1, 5.2, 5.3) and N columns for the N target gases. Among the first alternative options... - The generating device is configured to generate the matrix pseudo-inverse matrix and - The determiner (9) is configured to use the equation The N target gas concentrations are provided at its N outputs, and In the second alternative option The determiner (9) is configured to - Generate the matrix pseudo-inverse matrix and - Using the equation The N target gas concentrations are provided at its N outputs.

12. A method for generating a determiner (9) for generating a gas detection device (100) according to any one of claims 1 to 9, The system provides a training gas sample set, which includes multiple gas samples. The corresponding chemical composition of each gas sample in the training gas sample group is known, and the target gas to be detected (CH4, C3H8, C4H) is also known. 10 ) appears in at least one gas sample, and The generation method includes the following steps for each gas sample in the training gas sample group: - The gas sample is guided into the measuring chamber (2) of the gas detection device (100). - Each measurement detector (5.1, 5.2, 5.3) generates a signal, which is compared with the N target gases (CH4, C3H8, C4H) in the gas sample (Gp). 10 The corresponding concentration of ) is related, and - Generate training sample elements, the training sample elements including For each target gas to be detected, the target gas (CH4, C3H8, C4H) in the gas sample (Gp) 10 The concentration indication of ) and The identifier for each measurement detector (5.1, 5.2, 5.3) of the signal (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6) generated by that measurement detector (5.1, 5.2, 5.3) for that gas sample (Gp). The generation method described therein includes yet another step: The determiner (9) is trained by applying the learning method to the training samples. The training samples used for training include training sample elements generated using the gas samples from the training gas sample group.

13. The generation method according to claim 12, Its features Each gas sample in the training gas sample group contains only one target gas to be detected (CH4, C3H8, C4H). 10 ), and does not contain the target gas or any other gas to be detected, and For each training sample, the target gas to be detected (CH4, C3H8, C4H) 10 The corresponding identifier for the concentration of the gas sample is a first identifier for the target gas contained in the gas sample, and a second identifier for any other target gas.

14. The generation method according to claim 13, Its features The first identifier is 1, and the second identifier is 0.

15. The generation method according to any one of claims 12-14, Its features The step of generating corresponding sample elements for each gas sample in the training gas sample group includes the following steps: Generate vector sample elements as the training sample elements, such that the vector sample elements include vectors. b sum vector x , The vector b This includes M vector elements for the M signals (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6), which depend on the M signals (Sg.1, Sg.2, Sg.3) generated by the M measurement detectors (5.1, 5.2, 5.3) for the gas sample, and... The vector x Includes N vector elements that depend on the concentrations of the N target gases in the gas sample, and The step of training the determiner (9) by applying the learning method to the training samples Includes the following steps: generating a matrix So that for the elements of the vector sample, the product x With the vector b The deviation should be as small as possible. The matrix It includes M rows for the M measuring detectors (5.1, 5.2, 5.3) and N columns for the N target gases. Among the first alternative options... - The generation method includes generating the matrix. pseudo-inverse matrix Another step, and - The determiner (9) is configured to use the equation The N target gas concentrations are provided at its N outputs, and In the second alternative option The determiner (9) is configured to - Generate the matrix pseudo-inverse matrix and - Using the equation The N target gas concentrations are provided at its N outputs.

16. The generation method according to claim 15, Its features Each gas sample in the training gas sample group contains only one target gas to be detected (CH4, C3H8, C4H). 10 ), and does not contain the target gas or any other gas to be detected, and The matrix assigned to a specific target gas among the N target gases The columns are generated using those signals from the M measurement detectors (5.1, 5.2, 5.3), which are generated if a gas sample containing only the target gas is used.

17. A method for detecting N target gases (CH4, C3H8, C4H) in a gas sample (Gp). 10 Gas detection methods, Among them, there are N target gases to be detected (CH4, C3H8, C4H). 10 ) is predetermined, and where N is a predetermined number greater than or equal to 2. The gas detection method described herein is performed using a gas detection device (100). The gas detection device (100) used therein includes - Measurement chamber (2), - A detection arrangement (10) with M measurement detectors (5.1, 5.2, 5.3), and - An evaluation device (8) with a signal processing determinant (9), Where M is a predetermined number greater than or equal to 2. The detection arrangement (10) is configured such that If the gas sample (Gp) does not contain any of the target gas to be detected, each measurement detector generates a reference signal, and For each different chemical component of the gas sample (Gp), at least one measuring detector (5.1, 5.2, 5.3) generates a signal that deviates from the reference signal of the measuring detector (5.1, 5.2, 5.3). The determiner (9) includes -M inputs (E.1, E.2, E.3), meaning one input (E.1, E.2, E.3) for each measurement detector (5.1, 5.2, 5.3), and -N outputs (A.1, A.2, A.3), representing each target gas to be detected (CH4, C3H8, C4H). 10 One output, The gas detection method includes the following steps: - The gas sample (Gp) to be examined is guided into the measuring chamber (2). - Each measurement detector (5.1, 5.2, 5.3) generates a signal (Sg.1, Sg.2, Sg.3), which is compared with the N target gases (CH4, C3H8, C4H4) to be detected in the gas sample (Gp). 10 The concentration of at least one target gas in the mixture is related to the concentration of the target gas. - Apply signals (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6) to each of the M inputs (E.1, E.2, E.3) of the determiner (9), the signals depending on the signals (Sg.1, Sg.2, Sg.3) of the corresponding associated measurement detectors (5.1, 5.2, 5.3), and - Each of the N outputs (A.1, A.2, A.3) of the determiner (9) provides information about the corresponding associated target gas to be detected (CH4, C3H8, C4H). 10 The information regarding the concentration of ) In the initial training phase, a training gas sample set with multiple gas samples is provided. During the training phase, a generation method is employed. A training gas sample set containing multiple gas samples is provided for the generation method. The corresponding chemical composition of each gas sample in the training gas sample group is known, and the target gas to be detected (CH4, C3H8, C4H) is also known. 10 ) appears in at least one gas sample, and The generation method includes the following steps for each gas sample in the training gas sample group: - The gas sample is guided into the measuring chamber (2) of the gas detection device (100). - Each measurement detector (5.1, 5.2, 5.3) generates a corresponding signal, which is compared with the N target gases to be detected (CH4, C3H8, C4H). 10 The concentration of the gas sample (Gp) is related to the corresponding concentration of the gas, and - Generate training sample elements, wherein the training sample elements include: For each target gas to be detected, the target gas (CH4, C3H8, C4H) in the gas sample (Gp) 10 The concentration indication of ) and The identifier for each measurement detector (5.1, 5.2, 5.3) of the signal (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6) generated by that measurement detector (5.1, 5.2, 5.3) for that gas sample (Gp). The generation method described therein includes yet another step: The determiner (9) is trained by applying the learning method to the training samples. The training samples used for training include training sample elements generated using the gas samples from the training gas sample group.

18. The gas detection method according to claim 17, Its features M is a predetermined number that is greater than or equal to N.

19. The gas detection method according to claim 17 or 18, Its features The step of generating corresponding training sample elements for each gas sample in the training gas sample group includes the following steps: Generate vector sample elements as the training sample elements, such that the vector sample elements include vectors. b sum vector x , The vector b This includes M vector elements for the M signals (Sg.1 / Sg.6, Sg.2 / Sg.6, Sg.3 / Sg.6), which depend on the M signals (Sg.1, Sg.2, Sg.3) generated by the M measurement detectors (5.1, 5.2, 5.3) for the gas sample. The vector x Includes N vector elements that depend on the concentrations of the N target gases in the gas sample. The step of training the determiner (9) by applying the learning method to the training samples Includes the following steps: generating a matrix So that for the elements of the vector sample, the product x With the vector b The deviation should be as small as possible. The matrix It includes M rows for the M measuring detectors (5.1, 5.2, 5.3) and N columns for the N target gases. Among the first alternative options... - The generation method includes generating the matrix. pseudo-inverse matrix Another step, and - The determiner (9) is configured to use the equation x = A -1 b The N target gas concentrations are provided at its N outputs, and In the second alternative option The determiner (9) provides information about the N target gases (CH4, C3H8, C4H) at each of the N outputs. 10 The steps of obtaining the information on the N concentrations of ) Includes the following steps: the determiner (9) - Generate the matrix A pseudo-inverse matrix A -1 ,and - Using the equation x = A -1 b The N target gas concentrations are provided at its N outputs.

Images