Method for compensating a sensor
Patent Information
- Authority / Receiving Office
- EP · EP
- Patent Type
- Applications
- Current Assignee / Owner
- ENDRESS & HAUSER GMBH & CO KG
- Filing Date
- 2024-07-10
- Publication Date
- 2026-06-17
Smart Images
Figure EP2024069437_20022025_PF_FP_ABST
Abstract
Description
[0001] Method for compensating a sensor
[0002] The invention relates to a method for compensating a sensor and a sensor for determining measured values of at least one chemical and / or physical process variable.
[0003] Sensors, in particular pressure sensors, are used, for example, in process and automation technology and serve to monitor and / or determine at least one, for example chemical and / or physical, process variable of a medium.
[0004] The process variable to be determined by the sensor can be, for example, the pressure, fill level, flow rate, temperature, pH value, redox potential, or the conductivity of the respective medium. The various possible measurement principles underlying the determination of the process variable are known from the state of the art and will not be discussed further here. Sensors for measuring pressure are primarily designed as so-called absolute, relative, or differential pressure sensors.
[0005] Such sensors typically comprise a sensor unit that comes into contact with the process at least partially and / or at least temporarily, and an electronics unit that serves, for example, to detect, evaluate, and / or feed signals. The electronics unit of the sensor is typically arranged in a housing and additionally has at least one connection element for connecting the electronics unit to the sensor unit and / or an external unit. The connection element can be any electrical connection; a wireless connection is also possible. The electronics unit and the sensor unit of the sensor can be designed as separate units with separate housings or as a single unit with a shared housing.
[0006] Typically, the sensor unit does not determine and / or monitor the at least one process variable directly, but instead determines and / or monitors at least one measured variable from which the at least one process variable can be calculated by the electronics unit. In the case of pressure sensors, for example, a measuring diaphragm can be used, onto which the pressure of the process medium acts and which transmits the pressure to a piezoresistive element using a transmission medium. The effect of this pressure on the measuring diaphragm can be detected by piezoresistive resistors. The resistors are usually constructed as a Wheatstone bridge in order to generate an output voltage, in particular a diagonal voltage, which is ideally linear to the pressure exerted on the measuring diaphragm. In reality, however, this relationship is not linear and is also dependent on cross-influences.In the case of pressure sensors, for example, temperature plays a significant role as a cross-influence factor. Ideally, the application area of a pressure sensor should span the widest possible temperature and pressure range. Particularly at the edges of this application area, cross-influences become greater and the dependence on the measured variable becomes more non-linear. To enable the sensors to be used in a wide application area, e.g., across a broad pressure and temperature range, both of the aforementioned effects (non-linearity and cross-influence) must be eliminated through sensor compensation.
[0007] In the case of pressure sensors, several pressure and temperature levels are used to compensate for nonlinearities and cross-influences of the sensor output signal using mathematical modeling. At least one cross-influence variable is set, and at least one measured variable is determined depending on the cross-influence variable. Such cross-influence variables can be, as already mentioned, temperature. Furthermore, humidity can also have an undesirable cross-influence on the measured variable.
[0008] Compensation is a time-consuming process that results in high investment costs for the compensation system, high energy consumption, large space requirements in production and a high throughput time for the sensors.
[0009] The invention is therefore based on the object of demonstrating a possibility by which the scope of the compensation measurement can be shortened and thus the compensation time can be reduced.
[0010] The object is achieved according to the invention by the method according to claim 1 and the sensor according to claim 10. The method according to the invention for compensating a sensor, in particular a pressure sensor, which serves to determine measured values of at least one chemical and / or physical process variable of a medium, in particular a pressure, provides that the sensor is designed to determine measured values for the at least one process variable of the medium as well as at least one cross-influence variable affecting the measured variable and the method provides at least the following steps:
[0011] Defining a measuring sequence which is to serve for the compensation of the sensor, wherein the measuring sequence provides that predetermined setpoint values for at least the at least one process variable and the at least one cross-influence variable are approached;
[0012] Recording real measured values for the at least one process variable and the at least one cross-influence variable by the sensor as well as a sensor reference at the predetermined target values of the at least one process variable and the at least one cross-influence variable for at least part of the defined measuring sequence;
[0013] Interpolation of the real measured values recorded by the sensor using the real measured values recorded by the sensor reference to the specified target values of the at least one process variable and the at least one cross-influence variable;
[0014] Prediction of virtual measured values for the sensor using at least the actual measured values interpolated to the specified target values from the sensor for at least one other part of the specified measuring sequence in which no actual measured values were recorded;
[0015] Combining the actual measured values acquired by the sensor and the virtual measured values predicted for the sensor to form a characteristic field; determining compensation coefficients for a compensation equation of the sensor based on the characteristic field.
[0016] The invention is based on the consideration that the main factor in the time consumption during compensation is the operating time of the individual measuring stages or measuring points that must be approached during compensation. Thus, reducing the number of measuring stages or measuring points to be approached significantly shortens the compensation time. To this end, the invention proposes determining a portion of the actual measured values virtually.
[0017] An advantageous embodiment of the method according to the invention provides that, during interpolation, each acquired real measured value is interpolated over at least one, preferably several neighboring points, particularly preferably over all points in the characteristic curve field. In particular, the embodiment provides that the neighboring points and / or points of the characteristic curve field over which the interpolation is performed are actually acquired measured values of the at least one process variable and the at least one cross-influence variable.
[0018] A further advantageous embodiment of the method according to the invention provides that at least one interpolation polynomial, in particular at least one third-order interpolation polynomial is used for the interpolation of the at least one process variable and / or the at least one cross-influence variable.
[0019] A further advantageous embodiment of the method according to the invention provides that the virtual measured values are predicted from the interpolated actual measured values using a previously determined statistical model. In particular, the embodiment provides that, to determine model parameters for the statistical model, a sample of identical sensors is previously used to record actual measured values and / or to virtually predict measured values for the characteristic field intended to compensate the sensor.
[0020] A further advantageous embodiment of the method according to the invention provides that a multilinear, a multipolynomial model or a neural network is used as the statistical model.
[0021] A further advantageous embodiment of the method according to the invention provides that the compensation coefficients are made available, in particular stored or deposited, in the sensor for determining compensated measured values for a measuring operation.
[0022] A further advantageous embodiment of the method according to the invention provides that a standard deviation of the actually recorded measured values interpolated to the defined target values is also used for the prediction of the virtual measured values.
[0023] The invention further relates to a sensor, in particular a pressure sensor, for determining measured values of at least one chemical and / or physical process variable of a medium, in particular a pressure, wherein the sensor is designed to determine the measured values for the at least one process variable by means of a compensation equation with corresponding compensation coefficients, in particular with compensation coefficients which were determined according to the method according to claim 1.
[0024] The invention is explained in more detail with reference to the following drawings. It shows:
[0025] Fig. 1 : a schematic representation of a sensor to which the method according to the invention can be applied,
[0026] Fig. 2: a schematic process flow of the method according to the invention for compensating a sensor, as shown and described, for example, in Fig. 1, and
[0027] Fig. 3: a characteristic curve field with real measured values for the process variable and the cross-influence variable at a part of the defined setpoints,
[0028] Fig. 4: a characteristic field with both interpolated real measured values and virtually predicted measured values, and
[0029] Fig. 5: a characteristic field with the real measured values and the virtually predicted measured values.
[0030] Fig. 1 schematically shows a pressure sensor 2 which determines the pressure as a physical process variable of a medium 1. The medium 1 is arranged, for example, in a container 5. The sensor 2 has a sensor unit 3 and an electronics unit 4. In the example in Fig. 1, the sensor unit 3 and the electronics unit 4 are arranged in a common housing. Alternatively, separate housings can be provided for the sensor unit 3 and the electronics unit 4, or the two units can be arranged spatially separate from one another. The sensor unit 3 is designed to detect the physical process variable in the form of the pressure of the medium 1 as well as a cross-influence variable in the form of the temperature and to provide corresponding (analog or digital) measured values in the form of a signal to the electronics unit 4 for further processing.The electronics unit 4 is in turn designed to output a pressure measurement signal of the process variable compensated with respect to the cross-influence variable by means of a compensation equation with compensation coefficients which were determined according to the method to be described in more detail.
[0031] The method for compensating a sensor according to the invention can generally be used for all types of sensors. To clarify the method, it is described below using a pressure sensor as an example.
[0032] Fig. 2 shows a schematic process flow of the compensation method according to the invention.
[0033] In a first method step, a measurement sequence is defined which is to serve to compensate the sensor 2. For this purpose, a measuring field is shown as an example in Figs. 3 to 5 which comprises the process variable p, which is to be detected by the sensor 2, and the cross-influence variable T, which is also detected by the sensor 2. The measuring field 6 shown here and the method described below are based on a two-dimensional measuring field 6. The two-dimensional measuring field 6 comprises the process variable pressure (p) on the x-axis and the cross-influence variable temperature (T) on the y-axis. The invention is not restricted to a two-dimensional measuring field 6, but is also applicable to or transferable to an n-dimensional measuring field in which several process and / or cross-influence variables p. T are included.For example, humidity can be included as a further cross-influence variable, so that the measuring field 6 would be three-dimensional in this case.
[0034] In a second method step following the first, real measured values for the process variable p and the cross-influence variable T are recorded by the sensor and by a sensor reference. Real measured values are understood to be measured values recorded by the sensor 2 to be compensated itself and by the sensor reference. The real measured values are recorded at predetermined setpoints within the measuring field 6. The setpoints can be located at the intersection points of the vertical (cross-influence variable = temperature) and horizontal (process variable = pressure) lines, as shown in Figs. 3 to 5. In order to record the real measured values at the predetermined setpoints, the sensor 2 and the sensor reference are exposed to corresponding test conditions, i.e. in particular to a corresponding pressure and a corresponding temperature. The real measured values can be, for example, analog or digital raw measured values from the sensor unit of the sensor.
[0035] According to the invention, however, real measured values are not recorded by the sensor at all specified target values or intersection points within the characteristic curve field 6, but only at a portion of them. In the characteristic curve field shown in Fig. 3, for example, a total of only 15 real measured values are recorded at various target values by the sensor and the sensor reference. The actually recorded measured values are represented in Figs. 3 to 5 by circles with a solid line.
[0036] In a subsequent third method step, as shown in Fig. 3, the actual measured values recorded by sensor 2 for the process variable p and the cross-influence variable T at the specified setpoints are interpolated to the respective specified setpoints using the actual measured values recorded by the sensor reference. This preferably involves interpolating at least several neighboring points (the intersection points or setpoints adjacent to the respective setpoint or intersection point in the measuring field in the horizontal and / or vertical direction) of the actually recorded measured values. Particularly preferably, interpolation is carried out over all setpoints. For the interpolation of the actual measured values, an interpolation polynomial, for example, can be used; in this case, a 2D interpolation polynomial is used for each of the process variable pressure p and the cross-influence variable temperature T. The interpolation calculates the measured values at the defined setpoints, i.e.The actual measured values are corrected accordingly so that the interpolated (calculated) values are present at the defined target values. As a result, the actual measured values are converted to ideal measurement conditions. This is indicated in Fig. 3 by arrows for the respective measured value. The resulting interpolated measured values are temporarily stored for further use according to the method according to the invention.
[0037] Subsequently, in a fourth method step, as shown in Fig. 4, virtual measured values for the process variable p and the cross-influence variable T are predicted for the sensor. The interpolated measured values actually recorded by the sensor at the specified target values are used to help with this. In addition, the standard deviation of the actually recorded measured values interpolated to the specified target values can also be used to predict the virtual measured values. Based on these values (actually recorded measured values and / or standard deviations calculated for them), virtual measured values are predicted for at least part of the measuring field for which no actual measured values were recorded. Preferably, virtual measured values are predicted for all target values for which no actual measured values were recorded with the sensor. The virtually predicted measured values are represented in Figs. 4 and 5 by a circle with a dashed line.A previously defined statistical model can be used to predict the virtual measured values. The statistical model can be, for example, a multilinear, a multipolynomial model, or a neural network. The use of a multilinear or multipolynomial model offers the advantage that the model parameters can be determined very easily. A sample of identical sensors 2 can be used to create the model, with the help of which both actual measured values are recorded and virtual measured values are predicted over a defined period of time. During model creation and the associated parameterization, both the actual measured values recorded by the various sensors and the virtually predicted measured values are interpolated to the specified target values. The interpolation can be carried out in the manner described above.It is advantageous, but not absolutely necessary, to use all real measurement values recorded by the sensors to predict the virtual measurement values. The sample size, i.e., the number of sensors used to record the real measurement values for modeling, increases with the increasing number of model parameters.
[0038] Subsequently, in a fifth process step, as shown in Fig. 5, the actually recorded measured values and the virtually predicted measured values are combined to form a complete characteristic curve field 6. Preferably, the non-interpolated measured values actually recorded by the sensor are used here, since otherwise an unnecessary interpolation error would be included in the compensation.
[0039] In a subsequent sixth process step, compensation coefficients for a compensation equation used in the actual measuring operation of sensor 2 are determined based on the complete characteristic field 6. Using the compensation equation with the determined compensation coefficients, the process variable p is made independent of the cross-influence variable T during the actual measuring operation of sensor 2. Furthermore, the compensation equation can also be used to linearize the sensor signal for the process variable. In order to access the compensation coefficients during the actual measuring operation, they are ultimately made available in sensor 2. This can be achieved, for example, by storing them in the sensor, in particular in the sensor's electronics unit.
[0040] List of reference symbols
[0041] 1 medium
[0042] 2 Sensor, especially pressure sensor
[0043] 3 Sensor unit
[0044] 4 Electronic unit
[0045] 5 containers
[0046] 6 Characteristic field p Process variable, especially pressure
[0047] T Cross-influence variable, especially temperature
Claims
Patent claims 1 . Method for compensating a sensor (2), in particular a pressure sensor, which serves to determine measured values of at least one chemical and / or physical process variable (p) of a medium (1), in particular a pressure, wherein the sensor is designed to determine measured values for the at least one process variable of the medium (1) and at least one cross-influence variable (T) affecting the measured variable, and the method provides at least the following steps: Defining a measuring sequence which is to serve for the compensation of the sensor, wherein the measuring sequence provides that predetermined setpoint values for at least the at least one process variable and the at least one cross-influence variable are approached; Recording real measured values for the at least one process variable and the at least one cross-influence variable by the sensor as well as a sensor reference at the predetermined target values of the at least one process variable and the at least one cross-influence variable for at least part of the defined measuring sequence; Interpolation of the real measured values recorded by the sensor using the real measured values recorded by the sensor reference to the specified target values of the at least one process variable and the at least one cross-influence variable; Prediction of virtual measured values for the sensor using at least the actual measured values interpolated to the specified target values from the sensor for at least one other part of the specified measuring sequence in which no actual measured values were recorded; Combining the actual measured values acquired by the sensor and the virtual measured values predicted for the sensor to form a characteristic field; determining compensation coefficients for a compensation equation of the sensor based on the characteristic field.
2. Method according to claim 1, wherein during the interpolation, for each recorded real measured value, interpolation is carried out over at least one, preferably several neighboring points, particularly preferably over all points in the characteristic field.
3. Method according to the preceding claim, wherein the neighboring points and / or points of the characteristic field are interpolated over the, actually recorded measured values of the at least one process variable (p) and at least one cross-influence variable (T).
4. Method according to one or more of the preceding claims, wherein at least one interpolation polynomial, in particular at least one third-order interpolation polynomial is used for the interpolation of the at least one process variable (p) and / or the at least one cross-influence variable (T).
5. Method according to one or more of the preceding claims, wherein the virtual measured values are predicted from the interpolated real measured values using a previously determined statistical model.
6. Method according to the preceding claim, wherein, in order to determine model parameters for the statistical model, a sample of identical sensors is used beforehand to record real measured values and / or to virtually predict measured values for the characteristic field which is to serve for the compensation of the sensor.
7. Method according to one of the two preceding claims, wherein a multilinear, a multipolynomial model or a neural network is used as the statistical model.
8. Method according to one or more of the preceding claims, wherein the compensation coefficients are made available, in particular stored or deposited, in the sensor for determining compensated measured values for a measuring operation.
9. Method according to one or more of the preceding claims, wherein a standard deviation of the actually recorded measured values interpolated to the defined target values is further used for the prediction of the virtual measured values.
10. Sensor, in particular pressure sensor, for determining measured values of at least one chemical and / or physical process variable (p) of a medium (1), in particular a pressure, wherein the sensor is designed to determine the measured values for the at least one process variable by means of a compensation equation with corresponding compensation coefficients, in particular with compensation coefficients which were determined according to the method according to claim 1.