Method for detecting potential fault conditions on an FMCW-based level gauge

Abstract

Method for detecting a fault condition on an FMCW-based level gauge (1) used to measure the level (L) of a fill material (3) located in a container (2), comprising the following method steps: - Emitting a radar signal (S HF ), - Receipt of a response signal (E HF ), - Creation of a measurement signal (s ZF ) based on at least the response signal (E HF ), - Determination of a correlation coefficient (K) by means of a correlation of the measurement signal (s) ZF ) with a reference signal (s ref ), and - Detection of the error state, provided the correlation coefficient (K) has a predefined minimum value (K min ) falls below.

Description

[0001] The invention relates to a method for detecting a possible fault condition on an FMCW-based level measuring device, and to a level measuring device suitable for carrying out this method.

[0002] In automation technology, particularly in process automation, field devices are frequently used to detect and / or control process variables. Sensors are employed to detect these variables; these sensors are found, for example, in level gauges, flow meters, pressure and temperature gauges, pH / ORP meters, conductivity meters, and so on. They measure the relevant process variables, such as level, flow rate, pressure, temperature, pH value, ORP, or conductivity. A large number of these field devices are manufactured and distributed by Endress+Hauser.

[0003] Non-contact measuring methods have become established for measuring the fill level of contents in containers because they are robust and require little maintenance (the term "container" in this invention also includes open containers such as basins, lakes, or flowing waters). A further advantage of non-contact measuring methods is their ability to measure the fill level virtually continuously, i.e., with very high resolution. Accordingly, radar-based measuring methods are predominantly used for this purpose. An established measuring principle is the FMCW (Frequency Modulated Continuous Wave) measuring principle. It is based on the transmission of a continuous radar signal and the comparison of the response signal, which is reflected from the surface of the contents, with the frequency of the currently transmitted radar signal.The radar signal frequency lies within a defined frequency band around a standardized center frequency (fa). Typically, frequency bands in the 6 GHz, 26 GHz, or 79 GHz bands are used. A key characteristic of the FMCW method is that the transmission frequency is not constant but changes periodically within a frequency band. This change can be linear and exhibit a sawtooth or triangular waveform; however, a sinusoidal waveform can also be used, depending on the application.

[0004] A particular challenge with FMCW-based level measurement methods lies in the ability to unambiguously distinguish the received response signal, and the derived measurement signal, from superimposed interference. Ideally, the response signal should be generated solely by reflection of the radar signal from the surface of the contents. However, faulty response signals can be generated due to interference from various sources. A major cause is the reception of interference signals resulting from reflection of the radar signal off obstructions such as agitators or internal components within the tank. Internal interference signals can also occur, for example, if the source of the error is feedback within the antenna unit of the level gauge.

[0005] Numerous technical approaches now exist for filtering interference signals in FMCW-based level measurement in order to correct the response signal. For example, a method for calibrating FMCW-based level measuring instruments is known from the international publication WO 2012 / 139852 A1. This method uses a vibrating reference reflector positioned between the measuring instrument and the contents to generate a unique reference measurement signal, even during normal operation.

[0006] German patent application DE 10 2008 050 117 A1 describes a method for correcting internal interference signals of a level measuring device. The method described therein is based on measuring a reference signal in a largely absorbing test environment and subsequently creating a correction curve based on the reference signal.

[0007] While the aforementioned methods can compensate for interference signals, aging or contamination of the level gauge also alters the interference signals, or the measurement signal itself, over time. Therefore, it is advantageous to perform the correction during operation and to infer a fault condition from changes in the compensation values. Detecting such a fault condition is particularly desirable for level gauges used in critical process plants requiring high reliability. The necessary prerequisites for such applications are described, for example, in the IEC / EN 61508 standard for functional safety (also known as "Safety Integrity Level," or SIL).

[0008] The invention is therefore based on the objective of providing a method with which a fault condition in FMCW-based level measuring devices is reliably detected.

[0009] The invention solves this problem by a method for detecting a fault condition in an FMCW-based level gauge used to measure the fill level of a substance contained in a container. It comprises at least the following method steps: - Emitting a radar signal, - Receiving a response signal, - Creation of a measurement signal based on at least the response signal, - Determination of a correlation coefficient by correlating the measurement signal with a reference signal, and - Detection of the error state if the correlation coefficient falls below a predefined minimum value.

[0010] In this context, a fault condition is defined as a state of the level measuring device in which it is not ensured that the level measuring device determines a correct fill level.

[0011] Thus, the method according to the invention allows for the highly reliable detection of any fault condition in the level measuring device. This ensures that the level measuring device can also be used in process plants and measurement environments that require extremely reliable measuring instruments or measurement data.

[0012] Regarding the calculation of the correlation coefficient, it is advisable to use a cross-correlation, in particular according to the formula usually underlying cross-correlation. K=∑i=1n−φAZF,i*Aref,i+φ is calculated.

[0013] In the context of the invention, a reference signal is understood to be, in principle, any measurement signal that represents a defined reference situation for level measurement, either in the container or in another reference environment. Accordingly, there are various possibilities for implementing the reference signal: Firstly, the reference signal can be based on a theoretically derived ideal echo curve. For example, an ideally reflecting plane with a known distance and infinite extent can be used as the basis for this. Alternatively, the reference signal can be based on a reference measurement performed by the level measuring device under known reference measurement conditions in the container. For example, a prevailing or fallen below minimum fill level can be used as a reference measurement condition in the container. The presence of a clean container interior can also constitute an additional reference measurement condition.

[0014] Another possible reference measurement condition for creating a reference signal, in which only internal error sources of the level measuring device are represented, consists of recording the reference signal in a different test environment such as an absorption chamber, in which the radar signal is completely absorbed and thus no response signal is generated.

[0015] Furthermore, the inventive method can be further developed by generating a time-dependent function of the correlation coefficient based on the correlation coefficient and its change over continuous level measurements. Accordingly, if the correlation coefficient has not yet fallen below the predefined minimum value, a remaining operating time, during which the minimum value will be exceeded, can be calculated using the time-dependent function of the correlation coefficient. This further development is based on the idea of ​​approximating a remaining operating time, during which the minimum value is expected to be exceeded and thus the fault condition of the level measuring device will occur, by determining the temporal development of the correlation coefficient over at least two or more level measurements.This requires that the correlation coefficient has not yet exceeded the minimum value at the current time.

[0016] This further development of the invention allows for the prediction of a fault condition in advance, according to the principle of "predictive maintenance." The time-dependent function of the correlation coefficient can be generated using regression, in the simplest case a linear regression. Generally, the choice of a suitable regression type (i.e., exponential, logarithmic, polynomial, moving average, etc.) is not limited to a specific regression type, but rather depends on the individual time course of the correlation coefficient. Accordingly, the method of least squares, for example, can be used to perform the regression and / or to determine a suitable regression type.

[0017] The problem underlying the invention is solved analogously to the method according to the invention by a level measuring device for carrying out the method described in at least one of the variants described above. Such a level measuring device comprises: - A signal generation circuit for generating a radar-based signal, - a transmitting antenna for sending the radar signal, - a receiving antenna for receiving the response signal, - a mixer for generating the measurement signal by mixing the radar signal (S HF ) with the response signal, - an evaluation unit for ◯ Determination of the fill level based on the measurement signal, • Determination of the correlation coefficient based on the measurement signal and the reference signal, and for ◯ Detection of the error state if the correlation coefficient falls below a predefined minimum value.

[0018] It is therefore advisable to design the evaluation unit in such a way that it performs a Fourier transformation, in particular a Fast Fourier Transformation, of the measurement signal to determine the fill level, since this is computationally efficient compared to other calculation methods.

[0019] The invention is explained in more detail below with reference to the following figures. They show: Fig. 1: A standard arrangement of an FMCW-based level gauge on a container, Fig. 2: a typical circuit design of an FMCW-based level measuring device for carrying out the method according to the invention, and Fig. 3: Schematic representations for determining the correlation coefficient.

[0020] To understand the method according to the invention, in Fig. Figure 1 shows a typical arrangement of a level gauge 1 operating according to the FMCW measuring principle on a container 2. The container 2 holds a substance 3, the level L of which is to be determined by the level gauge 1. For this purpose, the level gauge 1 is mounted on the container 2 at a known installation height h above the substance 3. Depending on the application, the container 2 can be up to more than 100 m high.

[0021] The level measuring device 1 is arranged on the container 2 in such a way that it emits a radar signal S typical for FMCW. HF emits in the direction of the surface of the fill material 3. After reflection of the radar signal S HF The level measuring device 1 receives a corresponding response signal E at the surface of the contents (or, undesirably, at a disturbance object in the container 2, such as an inlet 21 protruding into the container). HF. As is characteristic of FMCW, the frequency difference between the currently emitted radar signal S HF and the response signal E HF This depends on the distance d = h - L to the surface of the contents. Therefore, the fill level can be determined based on the measured frequency difference.

[0022] The level sensor 1 is typically connected to a higher-level unit 4, such as a process control system, via a bus system, such as "PROFIBUS", "HART" or "Wireless HART". This allows information about any fault condition of the level sensor 1 to be communicated. Information about the fill level L can also be transmitted in order to control any inflows 21 and / or outflows 22 present at the tank 2.

[0023] In Fig. Figure 2 shows a suitable circuit design of an FMCW-based level measuring device 1, with which the inventive method for detecting a possible fault condition can be implemented: To generate a high-frequency signal typical of the FMCW measuring method s HF The level measuring device 1 includes a corresponding signal generation circuit 11, for example based on a PLL (phase-locked loop, i.e., a high-frequency oscillator kept frequency-stable by feedback, e.g., based on a voltage-controlled oscillator). The high-frequency signal s HFThe signal is designed to operate at a frequency in the microwave range (typically 6 GHz, 26 GHz, or 79 GHz, but frequencies up to over 100 GHz are also possible). This frequency is not constant but varies periodically within a predetermined frequency range: In the case of 79 GHz, the frequency range could be, for example, 2 GHz, resulting in a corresponding frequency between 78 GHz and 80 GHz. Typically, in FMCW (Functional Microwave Cirque) technology, this periodic change is a sawtooth-shaped (i.e., constant within this period) variation in the frequency of the radio frequency signal. HF It is about. However, any other form would also be conceivable, e.g. a sinusoidal change in frequency within the respective frequency difference.

[0024] The periodicity of the (sawtooth) change can, as is typical for the FMCW method, be on the order of up to several hundred MHz. The frequency difference of the high-frequency signal s HF In this case, the bandwidth should preferably be as large as possible, since increasing the bandwidth increases the resolution of the level measurement. Therefore, a generally higher frequency of the high-frequency signal is recommended. HF This is advantageous in terms of resolution because a higher absolute frequency difference can be implemented at higher frequencies.

[0025] After its generation, the high-frequency signal s HF The signal is fed to a transmitting antenna 14 via a signal divider 12 (and optionally a transmitting amplifier 13a). There, the electrical high-frequency signal s HF into the actual radar signal S HFThe signal is converted and transmitted accordingly. The design of the transmitting antenna is determined by the frequency of the radar signal S. HF made dependent. At frequencies between 6 GHz and 26 GHz, a horn antenna is often used. Especially at higher frequencies from 79 GHz, the antenna is usually designed as a planar antenna, e.g., a fractal, patch, or meandering antenna on the chip on which at least the RF components (11, 12, 13a / b, 14, 15, 16) of the level gauge 1 are located.

[0026] During measurement operation, the reflection of the radar signal S HF on the surface of the contents 3 (and / or on a foreign body in the container 2, such as an inlet 21 projecting into the container 2, see Fig. 1) a response signal E HF generated. In the case of a calibration or reference measurement on the level gauge 1, the response signal E results. HF from reflection of the radar signal S HFthrough appropriately predefined reference conditions, e.g., a reference object arranged in a measuring section at a known distance d. Another reference condition could also be defined by a precisely known fill level L in the container 2 itself (for example, with a known minimum fill level L). min , which, for example, cannot sink further due to a correspondingly arranged drain 22, see again Fig. 1) Furthermore, a largely reflection-free measurement environment (e.g., a suitable absorption chamber into which the level gauge is aligned) could also be considered as a reference condition. In this case, ideally no response signal E would be generated at all. HF The response signal E obtained under reference conditions HF This is then used as a reference signal. ref saved. Alternatively, to create the reference signal s refBy measuring under reference conditions, the reference signal could be determined. ref also based on a theoretically derived ideal echo curve.

[0027] The response signal E is received at a receiving antenna 15 of the level measuring device 1. HF The signal is received and converted back into an electrical signal (which can, if necessary, be amplified by a receiver amplifier 13b). This signal is then combined with the original high-frequency signal s by means of a receiver mixer 16. HF mixed, wherein the high-frequency signal generated by the signal generation circuit 11 s HF This is done by branching off from a signal divider 12. This results in a measurement signal typical of the FMCW method. ZF (also known as an intermediate frequency or IF signal) is generated. Its frequency f peakWithout interference, it depends solely on the distance d and thus enables the measurement of the fill level L. This contrasts with the frequencies of the radar signal S. HF and the response signal E HF is the frequency f peak of the measurement signal s ZF due to the difference formation between radar signal S HF and response signal E HF constant through mixer 16 (as long as the fill level L does not change).

[0028] Instead of a separate transmitting antenna 14 and a receiving antenna 15, a combined transmitting / receiving antenna could also be implemented using a suitable transmitting / receiving switch. At low frequencies up to approximately 26 GHz, this could be designed as a horn antenna, analogous to the use of two separate antennas, or as a planar antenna for higher frequencies.

[0029] To determine its frequency f peak(or, if the radar signal is also reflected by obstructions, a plurality of frequencies f peak ) the measurement signal s ZF The signal is typically subjected to analog-to-digital conversion by a digitization unit 17. The digitized measurement signal can then be used for... ZF The signal is subjected to a (near) Fourier transform and thus converted into an easily evaluable frequency spectrum. The frequency f can be determined from the maximum of the spectrum. peak and thus the distance d or fill level L can be determined. In the case of the Fig. In the embodiment shown in Figure 2, a corresponding evaluation unit 18 is provided for this purpose. If the evaluation unit 18 is designed, for example, as a microcontroller, FPGA or similar device, the digitization unit 17 can already be designed as an integral functional block of the evaluation unit 18.

[0030] The temporal profile of the measurement signal s ZF, i.e., before performing the Fourier transformation using the digitization unit 17, is in each case in Fig. 3a and Fig. 3b shown: Here, the [element] originates from the [element] shown. Fig. 3b shown measurement signal s ZF compared to the one in Fig. 3a plotted measurement signal s ZF a subsequent level measurement. From the comparison between Fig. 3a and Fig. 3b shows that the signal quality of the measurement signal s ZF such as the amplitude A or the frequency specificity can deteriorate with a successive number of measurements. Reasons for this could include, for example, the gradual build-up of dust on the transmitting antenna 14 and / or the receiving antenna 15 due to dusty material 3. Internal sources of error in the level measuring device 1, such as a misalignment of the mixer 17, could also be responsible.

[0031] The deterioration between the in Fig. 3a and Fig. 3b measured measurement signal s ZF Furthermore, in relation to a reference signal s ref , which is shown in both figures, is evident. The reference signal s ref for example, to use a stored measurement signal s ZF trade that was recorded under defined reference conditions. The one in Fig. The reference signal shown in 3a / b originates from the commissioning of the level measuring device 1, during which the inside of the container was clean and empty, i.e., the container 2 was filled to a maximum of the predefined minimum fill level L. min was filled with fill material 3.

[0032] The invention is based on the observation that a change in the measurement signal caused by interference influences s ZF very directly on the correlation coefficient K between the measurement signal s ZF and the previously stored reference signal s ref This has an effect. This becomes evident from the following: Fig. Figure 3c shows the course of the correlation coefficient K as a function of the increasing number N of level measurements. It shows that the correlation coefficient K weakens with increasing number N, provided that the signal quality of the measurement signal s decreases. ZF worsened by increasing disturbances.

[0033] The presence of a fault condition at the level measuring device 1 or a faulty level measurement can therefore be determined by defining a minimum value K. min The correlation coefficient K can be identified. Here, the minimum value K represents min represents a threshold value above which reliable level measurement is no longer possible and thus an error condition of the level measuring device 1 has occurred.

[0034] If the correlation coefficient K is determined using cross-correlation, the calculation can be based on the formula K=∑i=1n−φAZF,i*Aref,i+φ to be done. This involves A ZF,i and A ref,i the corresponding, phase-coherent and normalized amplitude values ​​of the measurement signal s ZF or the reference signal s ref (cf.) Fig. 3a and Fig. 3b). A prerequisite for the correct determination of the correlation coefficient K is that the measurement signal s ZF and the reference signal s ref , as especially in Fig. 3a shows that they do not exhibit any phase shift φ relative to each other. Otherwise, this must be corrected accordingly when calculating the correlation coefficient, as indicated in the formula.

[0035] Furthermore, as can be seen from the formula above, the value of the correlation coefficient K depends significantly on the amplitudes A of the measurement signal s. ZF and the reference signal s ref Regarding the minimum value K minTherefore, it is not possible per se to define a fixed value below which the level gauge 1 is to be classified as no longer functional. As from Fig. Since 3c can be derived, the minimum value K can be determined. min However, depending, for example, on the first measurement (N = 1, for example the measurement in which the reference signal s is also included). ref The predominant correlation coefficient K is defined (as determined), e.g., K. min = 1 / 10 *K N=1 A meaningful definition of the minimum value K min This can also be determined, for example, through appropriate aging tests on the level gauge.

[0036] If, during the ongoing operation of the level gauge 1, the minimum value K is reached min If the correlation coefficient K falls below a certain threshold, this is prevented by a logic gate 19 designed for this purpose (see Fig. 2) of the level measuring device 1 is detected. For this purpose, the correlation coefficient K between the measurement signal s is determined in the logic gate 19. ZF and the reference signal s ref e.g. calculated based on the formula described above.

[0037] To calculate the correlation coefficient K, logic gate 19 uses the measurement signal s. ZF at mixer 16. The reference signal s ref is received, for example, from evaluation unit 18 (see Fig. 2) The logic gate 19 itself can also be designed as an integral component of the evaluation unit 18. Any falling below the minimum value K min The presence of an error state can in turn be transmitted to the superior unit 4 via a corresponding fault message.

[0038] In Fig. Figure 3c also illustrates a further development of the method according to the invention. This further development is based on the idea of ​​determining a remaining operating time t by tracking the correlation coefficient K over at least two level measurements. r to approximate up to which the minimum value K is expected to be min The correlation coefficient K will fall below the minimum value required, thus triggering a fault condition at level gauge 1. To determine this, a time-dependent development of the correlation coefficient K, in the form of a corresponding mathematical function K(t), must be created based on the correlation coefficients K determined to date. Such a calculation is only meaningful, of course, if the correlation coefficient K at the time of the current level measurement meets the minimum value K. minhas not yet fallen below this threshold. The creation of the time-dependent function K(t) of the correlation coefficient K and / or the calculation of the remaining operating time t. r can in turn be done by the evaluation unit 18 (see Fig. 2) take place.

[0039] The in Fig. The time-dependent function K(t) shown in Figure 3c is based on regression of the changing correlation coefficient K using the moving average. In general, however, the choice of a suitable regression type (i.e., exponential, logarithmic, etc.) in accordance with the invention is not limited to a specific regression type, but rather should depend on the individual course of the correlation coefficient K (for example, the "least squares method" could be used to find a suitable regression type and / or to carry out the actual regression).

[0040] Following the creation of the time-dependent function K(t) of the correlation coefficient K, this function is used (starting from the correlation coefficient K of the measured signal s). ZF (at the time of the last level measurement) the expected remaining operating time t r approximates, where the correlation coefficient K has the minimum value K min The threshold will have been undercut. Therefore, by means of this further development of the invention, a fault condition can be detected in advance according to the principle of "predictive maintenance". Reference symbol list 1 level gauge 2 containers 3 Filling material 4. Higher-level unit 11 Signal generation circuit 12 signal dividers 13a, b Amplifier 14 transmitting antenna 15 Receiving antenna 16 mixers 17 Digitization circuit 18 evaluation units 19 logic gates 21 Inflow 22 Drain A Amplitude A ref,i Amplitude of the reference signal A ZF,i Amplitude of the measurement signal d distance E HF Response signal f frequency f peak Frequency of the intermediate frequency signal h Installation height of the level gauge i Indices of the individual measured values ​​of the measurement / reference signal K correlation coefficient K(t) Time-dependent function of the correlation coefficient K min Minimum value L level L min Minimum fill level N Number of measurements n Number of measured values ​​per measurement / reference signal S HF Radar signal s HF High-frequency signal s ref Reference signal s ZF Measurement signal t r Remaining operating time φ Phase shift of the reference signal

Claims

[1] Method for detecting a fault condition on an FMCW-based level measuring device (1) used to measure a level (L) of a fill material (3) located in a container (2), comprising the following method steps: - Emitting a radar signal (S HF ), - Receipt of a response signal (E HF ), - Creation of a measurement signal (s ZF ) based on at least the response signal (E HF ), - Determination of a correlation coefficient (K) by means of a correlation of the measurement signal (s) ZF ) with a reference signal (s ref ), and - Detection of the error state, provided the correlation coefficient (K) has a predefined minimum value (K min ) falls below. [2] Method according to claim 1, wherein the correlation coefficient (K) is determined by means of a cross-correlation, in particular according to the formula K=∑i=1n−φAZF,i*Aref,i+φ is calculated. [3] Method according to claim 1 or 2, wherein the reference signal is based on a theoretically derived ideal echo curve. [4] Method according to claim 1 or 2, wherein the reference signal (s ref ) is based on a reference measurement performed by the level measuring device (1) under previously known reference measurement conditions. [5] Method according to claim 4, wherein the reference measurement conditions are set when a minimum fill level (L) is undershot min ) predominate. [6] Method according to at least one of claims 1 to 5, wherein a time-dependent function (K(t)) of the correlation coefficient (K) is created based on the correlation coefficient (K) and its change over continuous level measurements, and where, in the event that the correlation coefficient (K) falls below the predefined minimum value (K min) does not fall below, based on the time-dependent function (K(t)) of the correlation coefficient (K) a remaining operating time (t) r ), to which the minimum value (K min ) is undershot, it is calculated. [7] Method according to claim 6, wherein the time-dependent function (K(t)) of the correlation coefficient (K) is determined by means of a regression. [8] Method according to claim 7, wherein the least squares method is used to perform the regression and / or to determine a suitable regression type. [9] Level measuring device for carrying out the method described in at least one of the preceding claims, comprising: - A signal generation circuit (11) for generating a radar signal (S HF ), - a transmitting antenna (14) for transmitting the radar signal (S HF ), - a receiving antenna (15) for receiving the response signal (E HF), - a mixer (16) for generating the measurement signal (s ZF ) by mixing the radar signal (S HF ) with the response signal (E HF ), - an evaluation unit (17, 18, 19) for ◯ Determination of the fill level (L) based on the measurement signal (s ZF ), ◯ Determination of the correlation coefficient (K) based on the measurement signal (s ZF ) and the reference signal (s ref ), and to ◯ Detection of the error state, provided the correlation coefficient (K) has a predefined minimum value (K min ) falls below.

10. Level measuring device according to claim 9, wherein the evaluation unit (17, 18, 19) is designed to determine the level (L) by performing a Fourier transform, in particular a Fast Fourier transform, of the measurement signal (s) ZF to carry out.

Images