EP1773172B1 - Method for calibrating sensors - Google Patents

Description

The invention relates to a method for calibrating sensors, in particular turbidity sensors in household appliances and an associated household appliance for carrying out the method.

In household appliances, eg. As dishwashers or washing machines, turbidity sensors for determining the degree of contamination of the cleaning liquid, eg. As washing liquor or rinse liquor used. With the aid of the values of the degree of soiling determined by the turbidity sensor, further control of the cleaning program of the household appliance takes place. In a dishwasher, for example, the cleaning program consists of the partial program steps "pre-rinsing", "cleaning", "intermediate rinsing", "rinsing" and "drying". Within the sub-program step "intermediate rinsing" often several intermediate rinsing steps are performed. By using the values of the degree of contamination determined by the turbidity sensor, it is possible to stop the execution of further intermediate rinsing steps by the controller of the dishwasher when it falls below a certain value of the degree of soiling. Thus, a considerable water and energy savings can be achieved with the same cleaning results. In addition, with a low degree of soiling during "pre-rinsing", the rinsing liquor from the "pre-rinse" can be used for the sub-program step "cleaning".

Turbidity is generally measured by passing light through the cleaning fluid. However, there are also other physical measuring methods, eg. B. conceivable with sound. When using the physical principle of passing light through the cleaning liquid, wherein particles in the cleaning liquid as a suspension retain a part of the light, a transmitting and receiving device for light is necessary. The transmitting device is, for example, a lamp or a light emitting diode and the receiving device z. B. to a phototransistor. However, the transceivers are subject to wear and aging changes. In addition, some significant deposits on the optical devices can occur. Temporary contamination at the transceivers can lead to significant errors in the measurements to lead. Over time, this leads to successively increasing errors in the measurements of the turbidity of the cleaning fluid. This leads to errors in the control of the household appliance.

From the EP 0 862 892 B1 is a household appliance with a measuring device for determining the degree of contamination of a cleaning liquid known. In order to prevent incorrect measurements, a comparison measurement with the measuring device in a cleaning program in which the measuring device is used to determine the degree of contamination of the cleaning liquid, previous cleaning program performed, preferably in a program part with unpolluted rinsing liquid, eg. B. rinsing, is performed. The measured value for the adjustment of the measuring device in the following cleaning program can be stored in a non-volatile memory. The disadvantage here is that with a small or hidden intermediate rinses not inconsiderable impurities in the rinsing liquor may also be contained during rinsing, so that the measurement results can be falsified. Furthermore, only one adjustment measurement is carried out, so that in the case of randomly occurring heavy soiling, eg. B. at the transmitting device by punctual deposits, measured values for the adjustment of the measuring device with significant errors are the result.

From the DE 101 11 006 A1 For example, a method for adjusting a turbidity sensor is known. Within a wash program, multiple calibration value measurements are performed at different times and stored in a first memory map, with calibration value measurements taken in multiple wash programs. From these calibration value measurements, the calibration measured value with the lowest degree of contamination is determined by selection for each wash program and written in a second memory table. From the stored, selected Kalibriermesswerten the second memory table, the average value is calculated, which forms the reference value for the measurement with the turbidity sensor.

Disadvantageously, only a relatively small number of calibration measurements form the basis for determining the reference value, which is only the mean value of a plurality of individual measurements within a wash program. This can be sources of error in the case of several washing programs or only within an entire washing program occur, for. B. contamination on the optics of the transmitting device, are not recognized. Due to the determination of the reference value by a mere averaging of all Kalibriermesswerten to a rinsing program these are loaded with frequently significant errors loaded calibration measured in the averaging disadvantageous. If there is contamination, for example, in the three preceding rinsing programs and this contamination is removed in a subsequent rinse program, nevertheless takes the measurement with the reference value from the mean of the frequently erroneous individual measurements, this error continues until all calibration measurements, the basis for the Form a reference value, are no longer subject to errors due to temporary impurities.

Object of the present invention is therefore to provide a method and an associated household appliance for performing the method, which allows a simple way under all operating conditions of a household appliance, especially for temporary contaminants, a reliable calibration of sensors, eg. B. turbidity sensors to allow.

This object is achieved by the method according to the invention for calibrating sensors according to claim 1. Advantageous developments of the invention are characterized by subclaims.

• Ermitteln von wenigstens zwei Messwerten in wenigstens einem Reinigungsprogrammablauf,
• Selektion wenigstens eines Messwertes durch Methoden der Statistik oder Wahrscheinlichkeitsrechnung, der im folgenden Schritt nicht mehr berücksichtigt wird,
• Ermittlung wenigstens eines möglichen Referenzwertes für die Kalibrierung des Sensors aus den nicht selektierten Messwerten und
• Selektion eines optimalen Referenzwertes aus wenigstens einem möglichen Referenzwert, sofern mehr als zwei mögliche Referenzwertes ermittelt wurden.

In the inventive method for calibrating a sensor, in particular a turbidity sensor in a household appliance, for. As a dishwasher or a washing machine, using reference values, the following steps are carried out:

• Determining at least two measured values in at least one cleaning program sequence,
• Selection of at least one measured value by methods of statistics or probability calculation, which is no longer considered in the following step,
• Determining at least one possible reference value for the calibration of the sensor from the non-selected measured values and
• Selection of an optimal reference value from at least one possible reference value if more than two possible reference values have been determined.

Expediently, the selection of the at least one measured value by methods of statistics or probability calculation, which is no longer considered in the following step, is carried out in each case from a series of measured values which were measured at the same time points within a rinsing program sequence. Thus, measured values are selected which were measured at the same times within a rinsing program sequence, so that they are similar to one another and suitable for further selection procedures or calculations.

• Ermitteln des arithmetischen Durchschnittes für die Messwerte gemäß der Formel $${\mathrm{d}}_{\mathrm{a}}=\frac{m_a^1+m_a^2+m_a^3+m_a^4+\dots +m_a^s}{s},$$
  
• Bestimmung des mittleren Fehlerquadrates nach der Formel $${{\mathrm{\sigma }}_{\mathrm{a}}}^2=\frac{{\left(m_1^1-d_a\right)}^2+{\left(m_1^2-d_a\right)}^2+\dots +{\left(m_1^s-d_a\right)}^2}{s},$$
  
• Ermittlung der wahrscheinlichen Grenzen des möglichen Referenzwertes, wobei diese innerhalb $$d_a\pm \frac{0,6746}{\sqrt{s}}$$
  
  liegen und
• Selektion der Messwerte, die außerhalb dieser Grenzen liegen.

Preferably, the following steps are carried out to select at least one measured value:

• Determining the arithmetic mean for the measured values according to the formula $${\mathrm{d}}_{\mathrm{a}}=\frac{m_a^1+m_a^2+m_a^3+m_a^4+...+m_a^s}{s}.$$
  
• Determination of the mean square error according to the formula $${{\mathrm{\sigma }}_{\mathrm{a}}}^2=\frac{{\left({\mathrm{<figure-callout\; id="1"\; label="m"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">m</figure-callout>}}_1^1-d_a\right)}^2+{\left({\mathrm{<figure-callout\; id="1"\; label="m"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">m</figure-callout>}}_1^2-d_a\right)}^2+...+{\left(m_1^s-d_a\right)}^2}{s}.$$
  
• Determination of the probable limits of the possible reference value, whereby these within $$d_a\pm \frac{0.6746}{\sqrt{s}}$$
  
  lie and
• Selection of measured values that are outside these limits.

In a further variant, if no measured value lies outside the probable limits of the possible reference value, the interval of the probable limits of the possible reference value is set smaller, so that at least one measured value lies outside and this selects at least one measured value. This is always at least a measured value is selected. The method can thereby be adapted to changing conditions.

In a further variant, predetermined empirical values are preferably additionally used from the factory to determine the probable limits of the possible reference value, which empirical values are automatically adapted in the course of the process to changing conditions. As a result, the method can be optimally applied to a new household appliance and there is an automatic adaptation to changing circumstances, eg. As impurities, so that the inventive method is "learning".

Preferably, the determination of the at least one possible reference value for the calibration of the sensor from the remaining, non-selected measured values is carried out by averaging. As a result, the possible reference values for the series of measured values for the measured values can be determined in a simple manner at any one time, and possibly existing incorrect measurements have only a small influence due to the averaging.

Expediently, the determination of the at least one possible reference value for the calibration of the sensor from the remaining, non-selected measured values is carried out by selecting a measured value by means of statistical or probability calculation methods. As a result, errors that result from individual possibly still existing incorrect measurements can be excluded, because only a single measured value is selected.

In a further variant, the measured value with the highest probability density is selected within the non-selected measured values. This makes it possible to rule out possible errors compared to averaging, which is based on measured values that may be subject to errors.

• Ermittlung der arithmetischen Mittelwerte der nicht selektierten Messwerte,
• Ermittlung des Betrages der Differenz aus dem arithmetischen Mittelwert und dem jeweiligen Messwert, wobei derjenige Messwerten ausgewählt wird, bei welchem der Betrag der Differenz am kleinsten ist.

Preferably, the measured value which is closest to the arithmetic mean of the non-selected measured values is selected as the reference value by the following steps:

• Determination of the arithmetic mean values of the non-selected measured values,
• Determining the amount of the difference from the arithmetic mean and the respective measured value, wherein that measured values is selected, in which the amount of the difference is the smallest.

In a supplementary variant, from the possible reference values the most optimal, ie generally the reference value with the smallest degree of contamination, is selected as the reference value for the calibration of the sensor.

Fig. 1

eine schematische Darstellung eines Trübungssensors,

Fig. 2

ein schematisches Ablaufdiagramm für ein Spülprogramm in einer Geschirrspülmaschine,

Fig. 3

ein erfindungsgemäßes Ablaufschema für die Ermittlung eines Referenzwertes für die Kalibrierung des Trübungssensors und

Fig. 4

ein weiteres erfindungsgemäßes Ablaufschema für die Ermittlung des Referenzwertes für die Kalibrierung des Trübungssensors.

The invention will now be described by way of example with reference to an embodiment with the aid of the following drawings:

Fig. 1

a schematic representation of a turbidity sensor,

Fig. 2

a schematic flow diagram for a washing program in a dishwasher,

Fig. 3

an inventive flowchart for the determination of a reference value for the calibration of the turbidity sensor and

Fig. 4

a further flow chart according to the invention for the determination of the reference value for the calibration of the turbidity sensor.

In Fig. 1 schematically a turbidity sensor 6 is shown. It has a transmitting device 1 as a lamp, which preferably emits visible light. The transmitting device 1 can also electromagnetic waves from other arbitrary frequency ranges, eg. As infrared light, emit. In a receiving device 2 as a photocell, the light incident on it is converted into electricity. Between the transmitting device 1 and the receiving device 2 is the rinsing liquor 3 with impurities. A control and evaluation unit 4 supplies the transmitting device 1 with power and evaluates the power supplied by the receiving device 2. The transmitting device 1 and the receiving device 2 are connected via electrical lines 5 to the control and evaluation unit 4. The control and evaluation unit 4 can also be part of the control of a dishwasher according to the invention, ie it is no separate control and evaluation unit 4 for the turbidity sensor 6 is necessary. Due to the change in the incident on the receiving unit 2 light at preferably constant power supply for the transmitting device 1, the degree of contamination of the wash liquor 3 is determined. The lower the power supplied by the receiving device 2, the greater the degree of contamination. The turbidity sensor 6 can in the dishwasher according to the invention z. B. be installed in the washing or in a line for rinsing. With the help of this value of the degree of contamination, the control of the dishwasher according to the invention controls the further program sequence. For example, falls below a certain degree of contamination, the implementation of further intermediate rinsing steps canceled or it takes place between pre-rinsing and cleaning no change in the rinse.

In Fig. 2 is a conventional Spülprogrammablauf s a dishwasher shown. On the abscissa axis the time is applied and on the ordinate axis the amount of rinsing liquor in the dishwasher. The wash program sequence consists of the program steps "Pre-wash", "Clean", "Intermediate wash", "Rinse" and "Dry". The one or more measured values m ₁ ¹ , m ₂ ¹ , m ₃ ¹ and m ₄ ¹ (m _a ^s with a as time t _{a of} the measurements within a Spülprogrammablaufes and s as the number of measurements of a measured value m _a ^s at the same time t _a in the wash program sequences s = 1, 2, 3 to s). For the calibration of the turbidity sensor 6, the measured values m ₁ ¹ , m ₂ ¹ , m ₃ ¹ and m ₄ ^{1 are} respectively determined at the times t _a = t ₁ , t ₂ , t ₃ and t ₄ . Within a rinsing program sequence s, only one measured value, preferably in the subprogram step "rinsing", can be measured, or also a plurality of reference values within the rinsing program, whereby also within a subprogram step, eg. B. "rinse", several readings, eg. B. m ₃ ¹ , m ₄ ¹ , for the calibration of the turbidity sensor 6 can be measured. The measured values m _a ^s at a time t _a are as a series of measurements in columns in Fig. 2 arranged one below the other. Within a rinsing program, the measured values are measured at the same time. There were according to Fig. 2 four measurements at different times t _a = t ₁ , t ₂ , t ₃ and t ₄ per wash program flow s performed, so that four columns of measurement series in Fig. 2 available. For different washing programs, eg. B. Soft 50 °, Intensive 70 ° or Automatic 55 ° -65 °, with different duration of the individual subprogram steps, the measurement is carried out at the same time after the beginning or before the end of a partial program step. In addition, the method can also be refined to the effect that a separate series of measurements is stored for each different rinse program with at least one measurement. In this procedure, the number of measurement series does not correspond to the number of measurement times t _a , but totals the sum of the individual measurement times t _a for each individual wash program.

In Fig. 3 an inventive flowchart for the determination of the reference value m _a ^{s is} shown. The top section shows the measured values m _a ^s . In a column in each case the measured values m _A ^s from the rinsing program sequences s = 1 are shown to s for each same time t _a. The Messerte m _a ^s are preferably slidably mounted on the respective current wash program sequence s + 1 preceding rinsing program sequences s = 1 to s determined. Here, another approach is possible, for. B. the measured values m _a ^s only from Spülprogrammabläufen s determined that have a low load, if z. B corresponding load sensors are present. The number of measurement times t _a thus corresponds to the number of columns. In the first column are z. At the time t _1, the measured values m _{a = 1} ^{s = 1 to s of} the rinsing program sequences s = 1 to s are listed, wherein measured values m _a ^s from different rinsing programs are also included herein.

als möglicher Referenzmesswert ermittelt. Aus diesen Mittelmesswerten $\overline{m_a^s}$

wird im darauffolgenden Operator der optimale Mittelmesswert $\overline{m_a^s}$

ausgewählt, welcher im Allgemeinen der Mittelmesswert $\overline{m_a^s}$

mit dem kleinsten Verschmutzungsgrad, d. h. der größte Mittelmesswert $\overline{m_a^s}$

ist. Dieser optimale Mittelmesswert $\overline{m_a^s}$

ist der Referenzwert für die Trübungsmessung im vorzugsweise darauffolgenden Spülprogramm. Neben dem Kriterium des kleinsten Verschmutzungsgrades können auch andere Kriterien, z. B. nur mögliche Referenzwerte aus einer bestimmten Spalte, wobei diese Kriterien auch ab Werk vorgegeben und/oder automatisch angepasst werden können.Below these columns is an operation unit. In this uppermost unit of operation, a selection is carried out at least by preferably statistical methods a measured value m _a ^s , which is no longer taken into account in the further steps. An example of such a statistical method will be described below. In addition to statistical methods, other methods come into consideration, eg. B. with methods of probability. Below the operation unit, the measured values m _a ^{s are} again arranged in columns from the rinsing sequences s, wherein in each case one measured value m _a ^{s was} selected. For example, in the 2nd column from the left of the measured value m ₂ ² from the rinse sequence was screened = s. 2 Below these columns is in Fig. 3 an operation unit shown. In the operation unit, the mean value of the remaining measured values m _a ^{s of} a column is formed, d, h. $\widetilde{m_a^s}$

determined as a possible reference measured value. From these mean readings $\widetilde{m_a^s}$

becomes the optimal mean reading in the following operator $\widetilde{m_a^s}$

which is generally the mean reading $\widetilde{m_a^s}$

with the smallest degree of contamination, ie the largest mean value $\widetilde{m_a^s}$

is. This optimal mean reading $\widetilde{m_a^s}$

is the reference value for the turbidity measurement in the preferably subsequent rinse program. In addition to the criterion of the smallest degree of contamination, other criteria, such. For example, only possible reference values from a specific column, these criteria can also be set ex works and / or automatically adjusted.

Alternatively to this procedure can be done accordingly Fig. 4 in this operation unit also from the selected measured values m _a ^s by selection method, eg. B. with methods of statistics, error theory or the probability calculation, a single measured value m * _a ^s from one column of measured values m _a ^s for each t _a , ie column, are selected. From these measured values m * _a ^s as possible reference values, the optimal measured value m * _a ^{s is} selected in the following operator, which is generally the measured value m * _a ^s with the smallest degree of contamination, ie the largest measured value m * _a ^s . This optimum measured value m * _a ^s is the reference value for the turbidity measurement in the preferably subsequent rinsing program.

The following is a statistical method for selecting at least one measured value m _a ^s corresponding to the uppermost operation unit in Fig. 3 and Fig. 4 described:

The measured values m _a ^s represent a number series m ₁ ¹ , m ₁ ² , m ₁ ³ , m ₁ ⁴ , m ₁ ⁵ ,..., M ₁ ^s for wash program sequences s for measured values at a time t _a . From these measured values m _a ^s , the arithmetic average d ₁ to d _a for measured values m _a ^s at the times t ₁ to t _{a is} determined with s as the number of measured values m _a ^s at a time t _a $${\mathrm{d}}_{\mathrm{a}}=\frac{m_a^1+m_a^2+m_a^3+m_a^4+...+m_a^s}{s}$$

Subsequently, the mean square error σ _a ^{2 is determined for} each a = 1, 2 to a. $${{\mathrm{\sigma }}_{\mathrm{a}}}^2=\frac{{\left({\mathrm{<figure-callout\; id="1"\; label="m"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">m</figure-callout>}}_1^1-d_a\right)}^2+{\left({\mathrm{<figure-callout\; id="1"\; label="m"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">m</figure-callout>}}_1^2-d_a\right)}^2+...+{\left(m_1^s-d_a\right)}^2}{s}$$

sind nach den Gesetzen der Fehlertheorie $$d_a\pm \frac{0,6746}{\sqrt{s}}$$

The probable limits of the reference value m * _a ^s or $\widetilde{m_a^s}$

are according to the laws of error theory $$d_a\pm \frac{0.6746}{\sqrt{s}}$$

Subsequently, it is checked in an algorithm whether measured values m _a ^s lie outside this probable limit. Are measured values m _a ^s outside, they will be selected. If there is an excessive number of measured values m _a ^s in relation to the number of measured values m _a ^s , only those measured values m _a ^s which are outside the probable limit by a specific value can be excluded. If there are no measurement values m _a ^s outside the border probable, metrics are m _a ^s exclude that lie at a certain value within the likely limitations. For this purpose, empirically determined values can also be specified ex works, which are preferably arithmetically adjusted in the course of the process to changing conditions.

This method is carried out for all rows of the measured values m _a ^s at the respective times t _a .

It is also possible, instead of the mean values d _a, to determine the probabilities of the individual measured values m _a ^s according to the probability calculation and to select those or those measured values m _a ^s with the least or the least probability, see Fig. 4 ,

In the following, a method of probability calculation for selecting a measured value m _a ^s according to the above second operation unit after Fig. 4 for use as a reference value for the calibration of a turbidity sensor from a series of measured values m _a ^s , z. B. m ₁ ¹ m ₁ ² , m ₁ ³ , m ₁ ⁴ , m ₁ ⁵ , ..., m ₁ ^s for Spülprogrammabläufe s explained.

According to the Gaussian hypothesis of the arithmetic mean, the probability density for the arithmetic mean of the measured value m _a ^{s is} greatest, regardless of how the Gaussian law of errors is designed.

From the remaining after the selection of at least one measured value m _a ^s measured values m _a ^s the arithmetic mean is determined d '. Subsequently, the distance of the individual measured values m _a ^s from this arithmetic mean d ' _a with d' ₁ , d ' ₂ , ... d' _{a is} to be determined, ie the magnitude | d ' _a - m _a ^s |. From these numbers, the smallest value is selected with an algorithm. The measured value m _a ^{s associated} with this smallest value is used as a possible reference value for the calibration of the turbidity sensor.

As a starting point for this selection of a measured value m _a ^s for use as a reference value, in a further variant of the method according to the invention, the measured values m _a ^{s can also be selected} before the selection in the uppermost operating unit of the at least one measured value m _a ^s . In the Fig. 4 existing uppermost operation unit is thus not used.

From these possible reference values m * _a ^s , whose number a corresponds to the number a of the times t _a for measuring the measured values m _a ^s within the rinsing program sequences s, the most optimal reference value is selected, which is generally the reference value with the smallest degree of contamination. This is done by means of a corresponding algorithm for determining the largest value.

Deviating from this procedure, the probability density can be determined for each measured value m _a ^s according to the laws of the probability calculation, and the measured value which has the greatest probability density can be selected as the reference value. The intermediate or final values determined in this method are preferably temporarily stored in non-volatile memories. The control is carried out with a corresponding computer system.

The present method according to the invention for calibrating sensors in household appliances makes it possible, by selecting individual measured values by statistical methods, to minimize errors resulting from the use of measured values with large deviations, for B. by temporary impurities, within a series of measurements to determine the reference value result. Individual measured values with large deviations are selected in particular using statistical methods.

The selection of a single measured value as a reference value, in particular with methods of the probability calculation, permits the error, which is determined by measured values with, in particular, strong deviations, caused by incorrect measurements, for example, as compared to an averaging. B. in the case of short-term deposits on the receiving or transmitting devices, arises to prevent.

Claims (10)

1. Method for calibrating a sensor, in particular a turbidity sensor (6) in a domestic appliance, e.g. a dishwasher or a washing machine, with the aid of reference values ( $\overline{m_a^s},$

   

   m*_a ^s) with the following steps:

   - establishing at least two measured values (m_a ^s) in at least one cleaning program sequence (s),

   - selection of at least one measured value (m_a ^s) by methods of statistics or probability calculation, which are not taken into further consideration in the following step and

   - establishing at least one possible reference value ( $\overline{m_a^s},$

   

   m*_a ^s) for the calibration of the sensor from the measured value (m_a ^s) not selected and

   - selection of an optimal reference value ( $\overline{m_a^s},$

   

   m*_a ^s) from the at least one possible reference value ( $\overline{m_a^s},$

   

   m*_a ^s), provided more than two possible reference values ( $\overline{m_a^s},$

   

   m*_a ^s) have been established.
2. Method according to claim 1, characterised in that the selection of the at least one measured value (m_a ^s) by methods of statistics or probability calculation, which are not taken into further consideration in the following step, is embodied from a series of measured values (m_a ^s) which were measured at the same points in time (t_a) within a rinse program sequence (s).
3. Method according to claim 1 or 2, characterised in that the following steps are carried out in order to select at least one measured value (m_a ^s):

   - establishing the arithmetic average (d₁ to d_a) for the measured values (m_a ^s for a= 1, 2, ..., a) according to the formula $${\mathit{d}}_{\mathit{a}}=\frac{m_a^1+m_a^2+m_a^3+m_a^4+\dots +m_a^s}{s},$$

   

   - determining the mean squared error (σ _a ² for (σ ₁ ² to (σ _a ²) with d_a from the first step according to the formula $${{\mathit{\sigma }}_{\mathit{a}}}^2=\frac{{\left(m_1^1-d_a\right)}^2+{\left(m_1^2-d_a\right)}^2+\dots +{\left(m_1^s-d_a\right)}^2}{s},$$

   

   - establishing the probable limits of the possible reference value (m*_a ^s, $\overline{m_a^s}$

   

   for m*₁ ^s, $\overline{m_1^s}$

   

   to m*_a ^s, $\overline{m_a^s}$

   

   ), wherein these lie between $$d_a\pm \frac{0,6746}{\sqrt{s}}$$

   

   and

   - selection of the measured values (m_a ^s), which lie outside said limits.
4. Method according to claim 3, characterised in that should no measured value (m_a ^s) lie outside the probable limits of the possible reference value (m*_a ^s, $\overline{m_a^s}$

   

   ), the interval of the probable limits of the possible reference value (m*_a ^s, $\overline{m_a^s}$

   

   ) is set smaller, so that at least one measured value (m_a ^s) lies outside and said at least one measured value (m_a ^s) is selected.
5. Method according to claim 4 or 5, characterised in that in order to establish the probable limits of the possible reference value (m*_a ^s, $\overline{m_a^s}$

   

   ), empirical values predefined ex works are preferably also employed, which are automatically adjusted to changing conditions in the method sequence.
6. Method according to claim 1 or 2, characterised in that the establishing of the at least one possible reference value $\left(\overline{m_a^s}\right)$

   

   for the calibration of the sensor from the remaining measured values (m_a ^s) not selected is carried out by averaging.
7. Method according to claim 1 or 2, characterised in that the establishing of the at least one possible reference value (m*_a ^s) for the calibration of the sensor from the remaining measured values (m_a ^s) not selected is performed by selection of a measured value (m_a ^s) by means of methods of statistics or probability calculation.
8. Method according to claim 7, characterised in that the measured value (m_a ^s) with the highest probability density within the measured values (m_a ^s) not selected is chosen.
9. Method according to claim 7 or 8, characterised in that that measured value (m_a ^s), which is closest to the arithmetic average value of the measured value m_a ^s not selected, is chosen as a possible reference value (m*_a ^s) by the following steps:

   - establishing the arithmetic average values (d'_a for a= 1, 2, a) of the measured values (m_a ^s) not selected

   - establishing the value of |d'_a - m_a ^s |, wherein that measured value (m_a ^s) is chosen, for which the value of | d'_a - m_a ^s | is smallest.
10. Method according to one of the preceding claims, characterised in that from the possible reference values ( $(\overline{m_a^s},$

    

    m*_a ^s), the most optimal, i.e. generally the reference value ( $(\overline{m_a^s},$

    

    m*_a ^s) with the smallest degree of contamination, is chosen as reference value ( $\overline{m_a^s},$

    

    m*_a ^s) for the calibration of the sensor.

Images