Method and apparatus for calibrating a gas probe for refrigerants

Abstract

A method for calibrating a gas probe (6) for a refrigerant, the gas probe (6) including at least two gas sensors (14a, 14b) each having a different sensitivity to a plurality of different refrigerants, the method comprising: (A) generating a gas mixture in a measurement chamber (3) including at least one gaseous refrigerant and at least one background gas; (B) detecting the gas mixture in the measurement chamber (3) using the gas probe (6) to be calibrated and storing measurement data provided by the gas sensors (14a, 14b) of the gas probe (6); (C) repeating steps (A) and (B) a plurality of times, the method including varying a concentration (c) of the at least one gaseous refrigerant and / or a concentration of the at least one background gas in the measurement chamber (3); and (D) using the measurement data provided by the gas sensors (14a, 14b) to calibrate an algorithm that enables identification of the type and concentration (c) of refrigerant contained in the gas mixture from the measurement data provided by the gas sensors (14a, 14b).

Description

[Technical Field]

[0001] The present invention relates to a method and a device for calibrating a gas probe for detecting refrigerants. The present invention also relates to a device for detecting refrigerants, comprising a gas probe calibrated by the method according to the invention and / or the device according to the invention. [Background technology]

[0002] Conventional technology Sensor systems are known that allow for the detection of refrigerant leakage from refrigeration and air conditioning systems, particularly air conditioning systems in motor vehicles. However, previously known sensor systems are unable to distinguish between different types of refrigerant. Summary of the Invention [Problem to be solved by the invention]

[0003] Disclosure of the Invention It is therefore an object of the present invention to provide a gas probe for detecting refrigerants, which makes it possible to determine not only the presence of refrigerant but also the type of refrigerant, and also to provide a method and device for calibrating such a gas probe. [Means for solving the problem]

[0004] The present invention includes a method for calibrating a gas probe for a refrigerant, the gas probe having at least two gas sensors with different sensitivities to a plurality of different refrigerant types. (A) generating a gas mixture in a measurement chamber, the gas mixture including at least one gaseous refrigerant and at least one background gas; (B) detecting the gas mixture in the measurement chamber using the gas probe to be calibrated and storing the measurement data provided by the gas sensor of the gas probe; (C) repeating steps (A) and (B) a plurality of times, the method including varying the concentration of at least one gaseous refrigerant and / or the concentration of at least one background gas in the measurement chamber before each repeat; (D) using the measurement data provided by the gas sensor to calibrate an algorithm that allows identifying the type and / or concentration of refrigerant contained in the gas mixture from the measurement data provided by the gas sensor; Includes.

[0005] The present invention also includes an apparatus for calibrating a gas probe for a refrigerant, the gas probe being provided with at least two gas sensors having different sensitivities to a plurality of different refrigerant types. a measuring chamber for containing the gas mixture and the gas probe to be calibrated; a gas metering device configured to meter a plurality of different gases, in particular a plurality of different types of refrigerant and / or a plurality of different background gases, into the measurement chamber; a control and evaluation unit connectable to the gas metering device and to the gas probe to be calibrated and adapted to carry out the method according to the invention for calibrating a gas probe, comprising the above-mentioned steps (A) to (D), Includes.

[0006] The device according to the invention may in particular comprise a random number generator, which makes it possible to randomly vary the concentration and / or type of at least one gaseous refrigerant in the measuring chamber and / or the concentration and composition of at least one background gas.

[0007] The concentration and / or type of at least one gaseous refrigerant in the measurement chamber and / or the concentration and composition of at least one background gas can be changed to a new one in each run, i.e., can be changed randomly based on a newly generated random number in each run.

[0008] Alternatively, a random number generator can be used to generate a random profile that emulates multiple leak searches, and this profile can then be used for the emulated leak searches without a new random number being generated each time. In this way, the computational cost of emulating the leak search can be reduced because the repeated generation of random numbers can be omitted.

[0009] The method according to the invention and the device according to the invention for calibrating a gas probe for refrigerants make it possible to provide a gas probe that can reliably distinguish between different refrigerants, for example, R134a and R1234yf. The method according to the invention and the device according to the invention also make it possible to increase the reliability of the calibration against potentially interfering background substances present in the detected gas mixture.

[0010] The present invention also includes a gas probe for detecting a refrigerant, comprising at least two gas sensors, each having a different sensitivity to a plurality of different refrigerant types and calibrated by the method for calibrating the gas probe according to the present invention, and an evaluation device configured to evaluate measurement data supplied by the gas sensors using an algorithm calibrated during calibration of the gas probe, thereby identifying the type of refrigerant detected.

[0011] The device may be particularly configured to distinguish between at least two different refrigerant types and / or to also identify the concentration of the detected refrigerant.

[0012] By using the device according to the invention for detecting refrigerants, it is possible to very efficiently and reliably identify the type and concentration of the detected refrigerant.

[0013] In one embodiment, the method for calibrating a gas probe includes repeating the steps of generating and detecting the gas mixture more than 10 times, particularly 20 to 100 times, to provide a sufficiently large data set for calibrating the algorithm, which can further improve the quality of the calibration.

[0014] In one embodiment, the method includes multiple increases and / or multiple decreases of the concentration of at least one gaseous refrigerant and / or the concentration of at least one background gas in the measurement chamber to generate measurement points that can be used to calibrate the algorithm. The method may particularly include stepwise or continuous changes of the concentration of at least one gaseous refrigerant and / or the concentration of at least one background gas in the measurement chamber.

[0015] The concentration of refrigerant in the measurement volume can be varied, for example, within a range from 0 ppm to 200 ppm.

[0016] If the range of interfering gas types and concentrations expected during typical use of the gas probe are known, these conditions can also be established when calibrating the gas probe, thus further improving the quality of the calibration.

[0017] If the range of interfering gases and concentrations expected during typical use of the gas probe is not known, standard parameters can be used for this purpose.

[0018] As organic background gases, for example, ethanol, acetone, and / or other hydrocarbons at concentrations between 0 ppm and 4 ppm can be used.

[0019] The method may also include holding the concentration of the at least one gaseous refrigerant and the concentration of the at least one background gas constant for a predetermined period of time.

[0020] In one embodiment, the method includes calculating the time derivative of the concentration of at least one gaseous refrigerant and / or the time derivative of the concentration of at least one background gas in the measurement chamber and taking these time derivatives into account when calibrating the algorithm. By taking the time derivatives of the concentrations into account, the quality of the calibration can be further improved.

[0021] In one embodiment, the method includes randomly varying the concentration of at least one gaseous refrigerant and / or the concentration of at least one background gas in the gas mixture.

[0022] In this case, the concentration of at least one gaseous refrigerant is changed continuously or quasi-continuously, in particular stepwise, for example by randomly varying the gradient of the refrigerant concentration over time within a defined window and adapting the refrigerant concentration accordingly.

[0023] In this way, when varying the concentration of at least one gaseous refrigerant and / or the concentration of at least one background gas in the gas mixture, the algorithm can be prevented from being trained on a fixed pattern and thereby identifying other patterns more poorly.

[0024] In one embodiment, the method comprises varying the concentration of at least one gaseous refrigerant and / or the concentration of at least one background gas in the gas mixture according to at least one predetermined profile, which may in particular comprise at least one profile corresponding to human behavior when searching for a refrigerant leak.

[0025] At least one predetermined profile may include a random component and / or may be randomly selected from a plurality of predetermined profiles, thus preventing the algorithm from being trained on a fixed set of profiles and consequently failing to identify different profiles.

[0026] In one embodiment, the method includes taking into account the humidity contained in the gas mixture when calibrating the algorithm, for example by measuring it using a humidity sensor in the gas probe. By taking into account the humidity contained in the gas mixture, the quality of the calibration can be further improved.

[0027] The method may also include varying the humidity content of the gas mixture to calibrate the algorithm with gas mixtures containing different humidity levels, thereby further improving the quality of the calibration.

[0028] In one embodiment, the method includes varying the type of refrigerant contained in the gas mixture between the different measurement sequences. The refrigerant contained in the gas mixture can be varied between the different measurement sequences, particularly randomly, to prevent the algorithm from being trained on a fixed refrigerant exchange pattern.

[0029] In one embodiment, the method comprises calibrating the algorithm using artificial intelligence techniques, which may in this case comprise, inter alia, the use of neural networks, decision trees and / or classical regression models, which allow for a particularly efficient and reliable calibration of the algorithm.

[0030] In one embodiment, the method comprises: using the first gas sensor to record, for each refrigerant type to be detected by the gas probe, a calibration curve representing the measurements provided by the first gas sensor as a function of the concentration of the refrigerant in the gas mixture, and forming an inverse of said function; For a second gas sensor, for a plurality of different gas mixtures containing a plurality of different background gases and a plurality of different background gas concentrations, carrying out the above-mentioned step (A) of generating a gas mixture containing at least one gaseous refrigerant and at least one background gas in a measurement chamber, and the above-mentioned step (B) of detecting the gas mixture generated in the measurement chamber using the gas probe to be calibrated and storing measurement data supplied from the gas sensor of the gas probe; performing a zero measurement in which no refrigerant is present in the gas mixture; correcting the measurement data provided by the second gas sensor by the result of the zero measurement; evaluating the frequency of the measurements as a function of the corrected measurements and defining a threshold therefrom, whereby measurements below the threshold are assigned to a first refrigerant type and measurements above the threshold are assigned to a second refrigerant type; Includes.

[0031] Such a method allows for the calibration of algorithms for identifying refrigerant type without using artificial intelligence techniques.

[0032] In the following, embodiments of the present invention will be described with reference to the accompanying drawings. [Brief explanation of the drawings]

[0033] [Figure 1A] 1 is a schematic diagram illustrating an apparatus for calibrating a gas probe for refrigerants according to one embodiment of the present invention. [Figure 1B] FIG. 1 is a schematic diagram illustrating a further apparatus for calibrating a gas probe for refrigerants according to an embodiment of the present invention. [Figure 2A] 1 is a schematic diagram illustrating a gas probe according to an embodiment of the present invention; [Figure 2B] FIG. 2 is a schematic diagram illustrating a further apparatus for calibrating a gas sensor for a refrigerant according to an embodiment of the present invention. [Figure 3A] FIG. 1 shows a diagram in which the concentration of a refrigerant is plotted as a function of time. [Figure 3B] 2 shows an exemplary progression of measurement data supplied by two gas sensors as a function of time t; FIG. [Figure 4A] FIG. 2 shows two calibration curves recorded for a first gas sensor. [Figure 4B] FIG. 10 shows a histogram in which the frequency of each measurement is plotted across the zeroed measurements of the second gas sensor. DETAILED DESCRIPTION OF THE INVENTION

[0034] DESCRIPTION OF THE DRAWINGS 1A shows a schematic diagram of an apparatus 1 for calibrating a gas probe 6 for a refrigerant according to one embodiment of the present invention. The apparatus 1 comprises a measurement chamber 2, which comprises a measurement volume 3 for containing a gas mixture.

[0035] The device 1 also includes a gas metering device 4 configured to meter a plurality of different gases, in particular a plurality of different types of gaseous refrigerants and / or a plurality of different background gases, into the measurement volume 3.

[0036] The measurement chamber 2 can be fitted with a gas probe 6 to be calibrated, which allows a portion of the gas mixture to be withdrawn from the measurement volume 3 in order to determine the composition of the gas mixture.

[0037] Optionally, a humidity sensor, not shown in FIG. 1A, can be provided in the measurement chamber 2, which measures the humidity of the gas mixture contained in the measurement volume 3.

[0038] The apparatus 1 further comprises a control and evaluation unit 8, which can be connected to the gas metering device 4, the gas probe 6 to be calibrated and, if desired, a humidity sensor. The control and evaluation unit 8 is configured to drive the gas metering device 4 and to receive measurement data from the gas probe 6 in order to be able to calibrate the gas probe 6. The control and evaluation unit 8 can comprise a random number generator 9, the function of which will be described below.

[0039] The gas metering device 4, the gas probe 6 and possibly the humidity sensor can each be connected via cables to a control and evaluation unit 8, as shown in FIG. 1A.

[0040] Alternatively, the gas metering device 4, the gas probe 6 and possibly the humidity sensor can each be connected to the control and evaluation unit 8 wirelessly, for example via a WiFi or Bluetooth® connection.

[0041] The gas probe 6 may be configured to first store the measured values ​​determined by the gas probe 6 and only transmit them wirelessly or via wire to the control and evaluation unit 8 once the measurement has been completed.

[0042] Excess gas introduced into the measuring volume 3 by the gas metering device 4 but not accommodated by the gas probe 6 can escape through a gas outlet 7 provided in the measuring chamber 2 .

[0043] Although only one gas probe 6 is shown in FIG. 1A, several gas probes 6 can be connected to the measurement chamber 2 and the control and evaluation unit 8 and calibrated simultaneously.

[0044] The measuring volume 3 of the measuring chamber 2 damps changes in the concentration of the gas introduced into the measuring chamber 2. This is particularly advantageous when the gas metering device 4 can only change the concentration of the gas introduced into the measuring chamber 2 in steps, but not continuously.

[0045] On the other hand, this damping effect of the measurement volume 3 limits the maximum possible temporal gradient of the gas concentration change. The damping effect can be varied by selecting the size of the measurement volume 3: the smaller the measurement volume 3, the smaller the damping effect, and vice versa. In an extreme case, the measurement chamber 2 with the smallest measurement volume 3 can be configured as a T-shaped element, as shown in FIG. 1B.

[0046] FIG. 2A shows a schematic diagram of a gas probe 6 constructed in accordance with the present invention.

[0047] The gas probe 6 includes a gas inlet 10, which may be configured at least partially as a flexible hose, so that the gas mixture to be detected can be easily taken from different locations, for example in the engine compartment of a motor vehicle.

[0048] The gas probe 6 may include a gas pumping device 12 for drawing gas through a gas inlet 10 , directing it through a gas sensor unit 14 within the gas probe 6 , and outputting it through a gas outlet 16 .

[0049] Alternatively, the gas probe 6 can be forced through by an externally blown-through flow with the gas to be measured. In this case, the gas probe 6 does not have the gas pumping device 12 or is provided with a switchable bypass that allows the gas pumping device 12 to be bypassed in order to prevent the forced through flow of the gas probe 6 from being blocked by the gas pumping device 12.

[0050] The gas sensor unit 14 includes at least two gas sensors 14a, 14b each having different characteristics, in particular, different sensitivities to different types of refrigerants that may be contained in the gas mixture to be detected.

[0051] The gas sensors 14a, 14b may be, for example, metal oxide gas sensors 14a, 14b having different characteristics, and in particular, gas sensors 14a, 14b provided by the manufacturer for detecting hydrocarbons, such as methane, may be used.

[0052] The gas sensors 14a and 14b can provide a measurement signal S in the form of an electrical resistance R or an electrical conductivity.

[0053] The gas sensors 14a, 14b may also provide as measurement signal S other physical quantities, for example a varying voltage, current or capacitance.

[0054] The gas sensors 14a, 14b may be configured with a filter layer 15a, 15b made of a chemical sorbent, such as activated carbon, at the inlet side. If such a filter layer 15a, 15b is not provided, it can be subsequently installed upstream of the respective gas sensors 14a, 14b.

[0055] The sensing characteristics of structurally identical gas sensors 14a, 14b can be varied by attaching different filter layers 15a, 15b to the gas sensors 14a, 14b. The characteristics of the gas sensors 14a, 14b and / or filter layers 15a, 15b can also be varied by operating the gas sensors 14a, 14b and / or filter layers 15a, 15b at different temperatures.

[0056] The gas probe 6 also includes a humidity sensor 5 configured to measure the humidity of the gas mixture flowing through the gas probe 6 .

[0057] The gas probe 6 also comprises a control unit 18 with a memory 24, which is arranged to drive the gas pumping device 12, to receive measurement data from the gas sensors 14a, 14b, to store and process them in the memory 24, and to output the measurement results via the data interface 20 and / or to display them on the display device 22. An acoustic output device 26 may also be provided, which is arranged to output an acoustic signal that depends on the measured refrigerant concentration c.

[0058] The gas sensors 14a, 14b can also be calibrated separately, i.e., without relying on the gas probe 6. For this purpose, the gas sensors 14a, 14b and the humidity sensor 5 are inserted into the measurement chamber 2 as shown in Fig. 2B. One or more gas sensors 14a, 14b can be inserted into the measurement chamber 2 simultaneously.

[0059] After calibration, the gas sensors 14 a, 14 b and the humidity sensor 5 are removed from the measurement chamber 2 and installed in the gas probe 6 .

[0060] FIG. 3A shows a diagram in which the concentration c of refrigerant in the measurement volume 3 is plotted on the vertical axis as a function of time t (horizontal axis).

[0061] The simulated measurement process typically involves a period of between 30 seconds and 30 minutes, in particular between 5 and 10 minutes.

[0062] To calibrate the gas sensor unit 14, the gas metering device 6 is driven by the control unit 18 so that the refrigerant concentration c in the measurement volume 3, which changes over time, simulates a search for a refrigerant leak by moving the gas probe 6 around a suspected refrigerant leak.

[0063] When the gas probe 6 is moved around the refrigerant leak, as the gas probe 6 approaches the refrigerant leak, the refrigerant concentration c detected by the gas probe 6 increases. As the gas probe 6 moves away from the refrigerant leak, the refrigerant concentration c measured by the gas probe 6 decreases.

[0064] The method according to the present invention described herein for calibrating the gas probe 6 does not involve moving the gas probe 6. Instead, the approach and receding of the gas probe 6 from a hypothetical refrigerant leak are simulated by varying the concentration c of refrigerant in the measurement volume 3 over time t, as exemplarily shown in Figure 3A.

[0065] The concentration c of the refrigerant in the measurement volume 3 can be varied, for example, within the range from 0 ppm to 200 ppm.

[0066] Before the first metering, the gas probe 6 is exposed to the gas for a period t0 of approximately 30 to 45 seconds, the length of which is intentionally chosen at random. In this case, a longer period t0 would be advantageous, but would result in a longer overall calibration period, so a compromise must be made here.

[0067] An additional interval of, for example, 5 to 10 minutes can be observed from the previous run, during which background gas is metered into the measuring chamber 2, but no coolant is yet metered in, so that the measured values ​​S supplied by the gas sensors 14a, 14b from the previous measurement cycle are attenuated as completely as possible.

[0068] The measurement values ​​S provided during this preceding decay phase are discarded. Optionally, the length of the decay phase can be flexibly set depending on the characteristics of the gas sensor 14 a, 14 b using a sensor signal stability criterion. In the case of a "perfect" or nearly "perfect" gas sensor 14 a, 14 b that does not exhibit substantial decay behavior, the additional decay phase can be omitted.

[0069] 3A, which illustrates the change in refrigerant concentration c with time t in measurement volume 3, includes regions 32 with positive slopes, in which the detected refrigerant concentration c increases with time. These regions 32 with positive slopes correspond to the approach of gas probe 6 to a refrigerant leak.

[0070] The graph also includes regions 34 with a negative slope, where the detected refrigerant concentration c decreases with time t. These regions 34 of decreasing concentration c correspond to situations where the gas probe 6 is moving away from the refrigerant leak.

[0071] 3A also includes a flat region ("plateau") 36 where the detected refrigerant concentration c is constant over time. If the gas probe 6 is not moved, or if the gas probe 6 is moved along a path where the refrigerant concentration is constant, the detected refrigerant concentration c will remain constant.

[0072] The hold duration during which the detected refrigerant concentration c is held constant may be selected randomly from a range between 5 seconds and 20 seconds, for example.

[0073] The gradient c'=dc(t) / dt and duration Δt of the individual regions 32, 34, 36 i may be specified within a given bandwidth, in particular by a random number generator 9 which may be provided in the control and evaluation unit 8, in order to simulate a random movement of the gas probe 6 around the refrigerant leak.

[0074] The slope c' may lie in the range between -100 ppm / min and +100 ppm / min. For another class of random profiles simulating smaller leaks that are difficult to find, a slope in the range between -25 ppm / min and +25 ppm / min can be chosen.

[0075] The random number generator 9 can also be used to simulate the behavior of a human user when searching for a refrigerant leak. When searching for a refrigerant leak, the user moves the gas probe 6, for example, along a path with increasing refrigerant concentration as long as possible. The random number generator 9 can therefore be used to ensure that the probability that the refrigerant concentration c will further increase along the simulated path is greater than the probability that it will decrease.

[0076] The generated profile, which represents the detected refrigerant concentration c as a function of time t, can be smoothed using a random number generator 9, thereby adapting the "sharp" changes in concentration that may result from an intended supply of gas into the measurement volume 3 to the rather "smooth" changes in concentration that would be detected if a person moved the gas probe 6 to search for a refrigerant leak.

[0077] Instead of using a random number generator 9 to simulate different movement profiles of the gas probe 6 around the gas leak, it is also possible to empirically identify the behavioral patterns of human users and from this generate movement profiles that approximate the typical behavior of a human user when searching for a refrigerant leak.

[0078] During a refrigerant leak detection simulation such as that shown in Figure 3A, a constant "background" of background gases, i.e., gases that are not constituents of the refrigerant, is established. This "background" does not change during the execution of a simulation such as that shown in Figure 3A.

[0079] In particular, during subsequent simulation runs, one other randomly selected background is set, and the change in the refrigerant concentration c over time t follows another randomly selected profile in each run.

[0080] The type of refrigerant introduced into the measurement volume 3 may also be randomly swapped between runs. If two refrigerant types are possible, the probability for each of these two possible refrigerant types may be taken as 50%.

[0081] If different probabilities for the presence of the two refrigerant types should be taken into account when the gas probe 6 to be calibrated is actually used, other probability distributions can also be used during calibration, for example a 20%:80%, 30%:70% or 40%:60% probability distribution, or an 80%:20%, 70%:30% or 60%:40% probability distribution.

[0082] In this case, the influence of prior and conditional probabilities should be taken into account to ensure that rarely used substances are discriminated to the same extent as frequently used substances. In particular, during calibration, rarely used substances should be measured approximately equally frequently.

[0083] Refrigerant mixtures containing two or more different refrigerant types in different proportions can also be used to train the algorithm to be able to recognize such mixtures.

[0084] Before each run, a new dosage of air humidity and organic matter is carried out into the measurement volume 3. The air humidity and the concentration of organic matter in the measurement volume 3 are kept constant during the run.

[0085] The air humidity may be varied, for example, between 30% and 70% at room temperature, i.e., between 20° C. and 25° C. As organic background gases, for example, ethanol, acetone, and / or other hydrocarbons can be used in concentrations between 0 ppm and 4 ppm.

[0086] The background gas concentrations may be defined, for example, by a random number generator or according to a statistical experimental design, where it is important to have a good coverage of the experimental space with a variety of different combinations.

[0087] When metering a time-dependent gas profile, it is possible that the gas metering device 4 does not allow for an exactly linear gas profile. Therefore, the gas concentration in the measurement volume 3 can be stepped in the form of a "step." Depending on the mixing of the gases in the measurement volume 3 before reaching the gas sensors 14a, 14b, this "step" can be "smoothed." For elaboration, this "smoothing" can also be modeled by calculation.

[0088] The time interval between the metering of background gas into the measurement volume 3 and the metering of refrigerant may be randomly defined, thereby preventing the algorithm from "memorizing" a fixed time interval between the metering of background gas into the measurement volume 3 and the metering of refrigerant and consequently making erroneous predictions based on the time elapsed between the start of the metering of background gas and the start of the metering of refrigerant.

[0089] After each emulated leak search, i.e., after a gas profile such as the one exemplarily shown in Figure 3A is run, it may also be advantageous to wait a predetermined minimum time for the sensor signal to decay before starting a new simulation run, especially if the previous run ended with a high refrigerant concentration c.

[0090] As an optional extension, in addition to the background gas in the measurement volume 3, further emulated disturbances that cause local variations in the refrigerant concentration c are considered, such as disturbances caused by wind or ventilation that result in a stronger dilution of the refrigerant concentration c, or emulated position-dependent disturbances.

[0091] FIG. 3B shows an exemplary progression S of the measurement data S supplied by the two gas sensors 14a, 14b. a ,S b as a function of time t.

[0092] The transition S of the measurement data S supplied from the two gas sensors 14a and 14b is caused by the difference in characteristics between the two gas sensors 14a and 14b. a ,S b are not congruent. Also, these transitions S a ,S b does not directly show the time course of gas metering as shown in FIG. 3A.

[0093] In the example shown in Figure 3B, the first gas sensor 14a is slower than the second gas sensor 14b. Therefore, the time course S of the sensor signal provided by the first gas sensor 14a a is the transition S supplied from the second gas sensor 14b. b and compared to the "real" refrigerant concentration produced by metering refrigerant into measurement volume 3 shown in FIG. 3A.

[0094] Furthermore, the progression S of the measurement signals supplied by the two gas sensors 14a, 14b a and S b are also significantly different from each other along the vertical axis, which may be due, for example, to the fact that the second gas sensor 14b responds differently to organic interfering gases than the first gas sensor 14a.

[0095] The measurement data S supplied by the two gas sensors 14a, 14b may differ both in their absolute value ("offset") and in their slope.

[0096] The two gas sensors 14a, 14b may have different response times, and therefore the signal transition S ais the signal transition S supplied by the other of the two gas sensors 14a, 14b. b are attenuated and / or delayed in time relative to

[0097] A humidity sensor 5 (see FIG. 1B) provided in the gas probe 6, which measures the humidity of the gas mixture flowing from the measurement volume 3 through the probe 6, provides a constant signal over time t.

[0098] From the measurements provided by the two gas sensors 14a, 14b, a number of time points t i Features are extracted in

[0099] The extracted features are used to train an algorithm that identifies the type of refrigerant in the gas mixture from measurements provided by the two gas sensors 14 a, 14 b. Optionally, the concentration c of the detected refrigerant in the gas mixture can also be identified.

[0100] The two gas sensors 14a and 14b are each provided with a signal at time t i The actual measurement value c(t i ), the extracted features are i the time derivative c'=dc(t) / dt of each sensor signal at t i -Δt), i.e., the previous time point t i It may also include the value of each sensor signal at -Δt. i -Δt may be, for example, 10, 20, 30, 40, 50 or 60 seconds in the past.

[0101] The concentration c(t) of the refrigerant produced in the measuring volume 3 by the gas metering device 4 and the type of refrigerant metered into the measuring volume 3 are known in this case and are calculated from the measured values ​​c(t) provided by the two gas sensors 14a, 14b. i ) and is used for training algorithms.

[0102] In contrast, the concentration of the background gas in the measurement volume 3 is assumed to be unknown and is therefore not used for training the algorithm.

[0103] Instead of the full cardinality information, a reduced summary of the information can also be used to train the algorithm. For example: - Only the type of refrigerant, without resolution of concentration c an increase and / or decrease in the concentration c, possibly with a threshold to be defined, and / or Semi-quantitative low / medium / high concentrations (with thresholds to be defined) can be used to train the algorithm.

[0104] In that case, the algorithm trained using the reduced information may predict from the sensor signal only the reduced information used to train the algorithm, instead of the refrigerant concentration.

[0105] As algorithms that are trained using the supplied measurement data, for example, neural networks, decision trees or classical regression models can be used.

[0106] Optionally, at the start of each measurement, the gas sensors 14a, 14b can be "zeroed," where a "baseline" based on the humidity of the gas mixture in the measurement volume 3 and the organic background is subtracted from the measurement.

[0107] Such a "zero setting" is particularly advantageous when the background gas is only reflected in the "baseline" and the sensitivity of the gas sensors 14a, 14b to the refrigerant gas remains unchanged. In this case, the humidity sensor 5 can be omitted, since the humidity contribution to the measurement result is already included in the "zero setting" offset.

[0108] If such a "zero setting" is carried out during calibration, a corresponding "zero setting" must also be carried out during operation of the gas probe 6 before the start of each measurement process.

[0109] This may be done manually, for example by pressing a "reset" button on the gas probe 6, or it may be automated. For example, the gas sensors 14a, 14b may measure a "zero value" immediately after the gas probe 6 is switched on, on which basis a measurement is performed and then the user is informed that the gas probe 6 is ready.

[0110] In the following, an alternative algorithm is described that can cope without the humidity sensor 5 in the gas probe 6 and without machine learning.

[0111] In the embodiment described here, the first gas sensor 14a is sensitive to two different refrigerants, e.g., R134a and R1234yf, as well as to humidity, and the second gas sensor 14b is sensitive only to humidity, organic (background) matter, and the second refrigerant, e.g., R1234yf, but not to the first refrigerant.

[0112] For each of the two refrigerants i=1, 2, a respective calibration curve S for the first gas sensor 14a is a (c i ) is recorded. FIG. 4A shows such a calibration curve S recorded for the first gas sensor 14a. a (c1),S a An example of (c2) is shown.

[0113] Two calibration curves a (c i ) for each inverse function c i (S a ) is formed, and its inverse function c i (S a ) is the concentration of each refrigerant c iis calculated by multiplying the measured value S a as a function of (i=1,2).

[0114] Advantageously, the gas mixture flowing from the measuring volume 3 through the gas probe 6 has an average relative humidity in the range between 40% and 60%, in particular in the range between 45% and 55%.

[0115] Then, the second gas sensor 14b is used to measure the varying refrigerant concentrations c for the two refrigerants (i=1, 2) as described above with reference to FIG. 3A. i Multiple “emulated leak searches” using (t) are performed within the measurement volume 3.

[0116] After the "zeroing" described above is performed, the emulated leak search is plotted in a histogram (see FIG. 4B), where the zeroed measurement result S b Each measurement result n(S b ) are plotted.

[0117] The histogram shows a bimodal distribution of the measurement results. In particular, the measurement results for a first refrigerant (i=1), e.g., R134a, are centered around 0 (zero), and the measurement results for a second refrigerant (i=2), e.g., R1234yf, are centered around the measurement value S of the second gas sensor 14b. b are widely distributed along the positive x-axis where

[0118] Here, the threshold S between the two distributions is th can be defined.

[0119] In a later measurement, the measured value S supplied by the second gas sensor 14b b is the threshold S th If it is below (S b th ), it is assumed that the first refrigerant (i=1) is detected.

[0120] ​ The measured value S supplied by the second gas sensor 14b b is the threshold S th If it exceeds (S b >S th ), it is assumed that the second refrigerant (i=2) is detected.

[0121] It should be noted that in this method, the presence of the first refrigerant (i=1) is only indirectly inferred from the non-detection of the second refrigerant (i=2). This may lead to errors when the refrigerant concentration is very low. Therefore, when only low refrigerant concentrations are detected, it may be advantageous to omit the explicit statement of the refrigerant type to avoid incorrect classification of the refrigerant type.

[0122] Concentration of each refrigerant c i is the measured value S supplied from the first gas sensor 14a a from the predefined inverse function c i (S a )

Claims

1. A method for calibrating a gas probe (6) for a refrigerant, the gas probe (6) including at least two gas sensors (14a, 14b) having different sensitivities to a plurality of different refrigerants, the method comprising: (A) generating a gas mixture in a measurement chamber (3) including at least one gaseous refrigerant and at least one background gas; (B) detecting the gas mixture in the measuring chamber (3) using the gas probe (6) to be calibrated and storing measurement data provided by the gas sensors (14a, 14b) of the gas probe (6); (C) repeating steps (A) and (B) multiple times, the method comprising varying the concentration (c) of the at least one gaseous refrigerant and / or the concentration of the at least one background gas in the measuring chamber (3); (D) using the measurement data provided by the gas sensors (14a, 14b) to calibrate an algorithm that allows identifying the type and concentration (c) of the refrigerant contained in the gas mixture from the measurement data provided by the gas sensors (14a, 14b); A method comprising:

2. The method comprises repeating steps (A) and (B) more than 10 times, in particular 20 to 100 times. The method of claim 1.

3. the method comprising increasing and / or decreasing the concentration (c) of the at least one gaseous refrigerant and / or the concentration of the at least one background gas in the measuring chamber (3), 3. The method according to claim 1 or 2.

4. the method comprising varying the concentration (c) of the at least one gaseous refrigerant and / or the concentration of the at least one background gas in the measuring chamber (3) in a stepwise or continuous manner; 4. The method according to any one of claims 1 to 3.

5. the method comprising taking into account the time derivative (c') of the concentration (c) of the at least one gaseous refrigerant in the measuring chamber (3) when calibrating the algorithm; 5. The method according to any one of claims 1 to 4.

6. the method includes measuring humidity of the gas mixture and taking it into account when calibrating the algorithm; The method comprises, in particular, varying the humidity of the gas mixture.

6. The method according to any one of claims 1 to 5.

7. The method includes performing a zero measurement in which no refrigerant is present in the gas mixture.

7. The method according to any one of claims 1 to 6.

8. the method comprises varying, in particular randomly varying, the type of refrigerant contained in the gas mixture; 8. The method according to any one of claims 1 to 7.

9. the method comprising randomly varying the concentration (c) of the at least one gaseous refrigerant and / or the concentration of the at least one background gas in the gas mixture; 9. The method according to any one of claims 1 to 8.

10. the method comprising varying the concentration (c) of the at least one gaseous refrigerant and / or the concentration of the at least one background gas in the gas mixture according to at least one predetermined profile; the at least one predetermined profile in particular comprises a random component; 9. The method according to any one of claims 1 to 8.

11. the method comprising generating a new random number at each execution; or the method including, in each execution, accessing a set of pre-generated random numbers; 11. The method according to any one of claims 1 to 10.

12. the method includes calibrating the algorithm using artificial intelligence techniques; The method includes, inter alia, using neural networks, decision trees, and / or classical regression models.

12. The method according to any one of claims 1 to 11.

13. The method comprises: Using a first gas sensor (14a), for each refrigerant type to be detected by the gas probe (6), the measurement value provided by the first gas sensor (14a) is converted into the concentration (c) of the refrigerant in the gas mixture. ｉ ) as a function of the calibration curve (S ａ (c ｉ )) and forming an inverse function of said function; performing steps (A) and (B) for a plurality of different gas mixtures including a plurality of different background gases and a plurality of different background gas concentrations for a second gas sensor (14b); performing a zero measurement in which no refrigerant is present in the gas mixture; The measurement data (S ｂ ) by the result of the zero measurement; The frequency (n) of the measurement results is calculated by the corrected measurement data (S ｂ ) as a function of , from which a threshold is defined (S ｔｈ ), thereby the threshold value (S ｔｈ ) below the measured value (S ｂ ) to the first refrigerant type, and the threshold value (S ｔｈ ) ｂ ) to a second refrigerant type; 12. The method of claim 1, comprising:

14. An apparatus (2) for calibrating a gas probe (6) for a refrigerant, the gas probe (6) comprising at least two gas sensors (14a, 14b) each having a different sensitivity, the apparatus (2) comprising: a measuring chamber (3) for containing the gas mixture and the gas probe (6) to be calibrated; a gas metering device (4) configured to meter a plurality of different gases, in particular a plurality of different refrigerant types and / or a plurality of different background gases, into the measuring chamber (3); a control and evaluation unit (8) connectable to the gas metering device (4) and to the gas probe (6) to be calibrated and configured to perform the method for calibrating a gas probe (6) according to any one of claims 1 to 12; An apparatus (2).

15. A gas probe (6) for detecting a refrigerant, said gas probe (6) comprising: at least two gas sensors (14a, 14b) with different sensitivities, calibrated by a method for calibrating a gas probe (6) according to any one of claims 1 to 12; an evaluation device (18) configured to evaluate the measurement data supplied by the gas sensors (14a, 14b) using an algorithm calibrated during calibration of the gas probe (6), thereby identifying the type and / or concentration of the refrigerant detected; Equipped with The evaluation device (18) is particularly adapted to distinguish between at least two different refrigerant types, The evaluation device (18) is configured in particular to determine the concentration (c) of the refrigerant detected. Gas probe (6).

Images