Fault recognition system for radar level gauge for volatile substances

Abstract

The application discloses a fault identification system for a radar liquid level meter for volatile substances and belongs to the technical field of radar liquid level meters. The system is used for solving the problems of fault omission and long fault checking process in the detection and analysis of the faults of the radar liquid level meter. The system comprises the following modules: a return signal extraction module, which adopts two-in wavelet transform to perform wavelet decomposition on the return signal of the radar liquid level meter, adopts principal component analysis to perform decorrelation, adopts a Fisher linear classification function to perform feature extraction, utilizes a BP neural network classifier to classify the features, and outputs a classification result; a return signal fitting module, which is used for drawing a fitting curve of the classification result output by the return signal extraction module and the return time of the radar liquid level meter, and obtaining a return graph of the return curves of different classification results; and a return graph analysis module, which is used for judging whether there is crystallization attachment on the antenna according to the fluctuation value on the return graph. The application is used for detecting the organic crystallization on the horn mouth or the antenna of the radar liquid level meter.

Description

TECHNICAL FIELD

[0001] The application relates to a fault identification system of a radar liquid level meter. BACKGROUND

[0002] The principle of the radar liquid level meter is that a probe sends out a high-frequency pulse, when the pulse meets the surface of a material, is reflected back to a receiver in the instrument, and is converted into a material level signal. The radar liquid level meter is generally used for non-contact continuous measurement of the material level of liquid, slurry and granular materials. Since the radar liquid level meter does not need to be in contact with the liquid to be measured during use, the problem that the detection element of the plug-in instrument is easily polluted and corroded by the medium can be solved; and the radar liquid level meter can also penetrate through foams, smokes, vapors and other media, and the measurement result is not affected by the fluctuating medium performance, nor by the process condition change. Therefore, the radar liquid level meter has been widely applied.

[0003] However, the main components of the radar liquid level meter are electronic elements and an antenna, and if there are volatile organic substances in the tank during use, the volatile organic substances will crystallize on the horn mouth or the antenna of the radar liquid level meter, which will cause interference waves during use of the radar liquid level meter, affecting the measurement accuracy.

[0004] Now, the detection and analysis of the faults of the radar liquid level meter mainly rely on the experience of the staff, which will lead to the omission of faults due to insufficient experience of the staff, and the process of finding and troubleshooting the faults is relatively long, thereby leading to the inaccuracy of the overall result of the continuous measurement process of the radar liquid level meter. SUMMARY

[0005] The application aims to solve the problems of fault omission and long troubleshooting process in the detection and analysis of the faults of the radar liquid level meter, and provides a fault identification system for the radar liquid level meter of volatile substances.

[0006] The fault identification system for the radar liquid level meter of volatile substances comprises an echo signal extraction module, an echo signal fitting module and an echo figure analysis module.

[0007] The echo signal extraction module adopts two-in wavelet transform to perform wavelet decomposition on the echo signal of the radar liquid level meter, adopts principal component analysis for decorrelation, adopts a Fisher linear classification function for feature extraction, uses a BP neural network classifier to classify the features, and outputs a classification result.

[0008] The echo signal fitting module is used for drawing a fitting curve of the classification result output by the echo signal extraction module and the echo time of the radar liquid level meter, and obtaining an echo figure of the echo curve of different classification results.

[0009] The echo figure analysis module is used to determine whether there is crystalline attachment on the antenna according to the fluctuation value on the echo figure.

[0010] Preferably, the specific method for wavelet decomposition of the echo signal of the radar level gauge using the dyadic wavelet transform comprises: mapping the echo signal of the radar level gauge into wavelet functions of different scales to obtain small-scale information and large-scale information of the echo signal respectively.

[0011] Preferably, the specific method for decorrelation using principal component analysis comprises:

[0012] S3-1, converting the small-scale information and the large-scale information into input data vectors respectively;

[0013] S3-2, obtaining a correlation matrix of the small-scale information input data vector, and then obtaining eigenvalues and eigenvectors of the small-scale information;

[0014] obtaining a correlation matrix of the large-scale information input data vector, and then obtaining eigenvalues and eigenvectors of the large-scale information;

[0015] S3-3, selecting M dominant eigenvalues, and eigenvectors corresponding to the M dominant eigenvalues constituting a feature subspace; the M is a positive integer;

[0016] S3-4, projecting the eigenvectors of the small-scale information obtained in S3-2 into the feature subspace obtained in S3-3;

[0017] projecting the eigenvectors of the large-scale information obtained in S3-2 into the feature subspace obtained in S3-3;

[0018] S3-5, completing decorrelation and compression of the high-dimensional data using principal component analysis.

[0019] Preferably, the specific method for feature extraction using the Fisher linear classification function comprises:

[0020] arranging the data after decorrelation and compression in descending order using the Fisher linear classifier;

[0021] calculating classification errors of the Fisher linear classifiers of different data respectively;

[0022] selecting a value with the largest class discrimination ability as the optimal feature subset according to the criterion of minimum classification error, and inputting the value to a neural network classifier.

[0023] Preferably, the specific method for classification of features using the BP neural network classifier comprises: using a feedforward multilayer neural network to classify the input data.

[0024] Preferably, the feedforward multilayer neural network uses a back propagation learning algorithm.

[0025] Preferably, the specific method for determining whether there is crystalline attachment on the antenna according to the fluctuation value on the echo map includes: when one of the echo curves appears a fluctuation value and the rest of the echo curves are smooth fluctuation, it is determined that there is crystalline attachment on the antenna.

[0026] Preferably, the fluctuation value appearing on the echo curve is in the range of 1 / 3 to 2 / 3 of the peak value.

[0027] Preferably, the smooth fluctuation of the echo curve is in the range of less than 1 / 4 of the peak value.

[0028] Preferably, the fluctuation value on the echo curve in the range of more than 2 / 3 of the peak value is the actual liquid level.

[0029] The advantages of the present application: the fault identification system for the radar liquid level meter for volatile substances proposed in the present application, the signal reflected by the radar liquid level meter is reflected on the echo map through the echo signal extraction module and the echo signal fitting module, the echo map can analyze the echo signal and determine whether there is crystalline attachment on the antenna. The crystallization of volatile substances has uncertainty, and the fault identification system of the present application can effectively and timely determine the crystalline attachment, so that the staff can timely find the crystallization of volatile substances and clean the crystallization, thereby minimizing the accuracy of continuous measurement of the radar liquid level meter. BRIEF DESCRIPTION OF DRAWINGS

[0030] Figure 1 is the principle block diagram of the fault identification system for the radar liquid level meter for volatile substances described in the present application;

[0031] Figure 2 is the echo map of the echo curve of different classification results simulated by taking the pulse transmission time interval of the radar liquid level meter as 1s and the frequency as 2.5MHz, wherein the horizontal coordinate is the echo time, the unit is s, and the vertical coordinate is the signal intensity, the unit is mV. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0033] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0034] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0035] Example 1:

[0036] The following is combined with Figure 1 This embodiment describes a fault identification system for radar level gauges used for volatile substances, which includes: an echo signal extraction module, an echo signal fitting module, and an echo graph analysis module.

[0037] The echo signal extraction module uses binary wavelet transform to perform wavelet decomposition on the echo signal of the radar level gauge, principal component analysis for decorrelation, Fisher linear classification function for feature extraction, and BP neural network classifier for feature classification, outputting the classification result.

[0038] The echo signal fitting module is used to plot the classification results output by the echo signal extraction module and the fitting curve of the radar level gauge echo time, and obtain the echo map of the echo curve of different classification results.

[0039] The echo image analysis module is used to determine whether there are crystalline deposits on the antenna based on the fluctuation values ​​on the echo image.

[0040] Furthermore, the specific method for wavelet decomposition of the echo signal of the radar level gauge using binary wavelet transform includes mapping the echo signal of the radar level gauge to wavelet functions of different scales to obtain the small-scale and large-scale information of the echo signal respectively.

[0041] Furthermore, specific methods for decorrelation using principal component analysis include:

[0042] S3-1. Convert the small-scale information and large-scale information into input data vectors respectively;

[0043] S3-2. Obtain the correlation matrix of the small-scale information input data vector, and then obtain the eigenvalues ​​and eigenvectors of the small-scale information;

[0044] Obtain the correlation matrix of the large-scale information input data vector, and then obtain the eigenvalues ​​and eigenvectors of the large-scale information;

[0045] S3-3. Select M dominant feature values, and the feature vectors corresponding to the M dominant feature values ​​constitute the feature subspace; where M is a positive integer;

[0046] S3-4. Project the feature vectors of the small-scale information obtained in S3-2 onto the feature subspace obtained in S3-3;

[0047] Project the feature vectors of the large-scale information obtained in S3-2 onto the feature subspace obtained in S3-3;

[0048] Principal component analysis was used to decorrelate and compress high-dimensional data.

[0049] Furthermore, specific methods for feature extraction using the Fisher linear classification function include:

[0050] Fisher linear classifier was used to sort the decorrelated and compressed data in descending order;

[0051] Calculate the classification error of the Fisher linear classifier for different datasets;

[0052] Based on the criterion of minimizing classification error, the values ​​with the greatest class discrimination ability are selected as the optimal feature subset and input into the neural network classifier.

[0053] Furthermore, specific methods for classifying features using a BP neural network classifier include: using a feedforward multilayer neural network to classify the input data.

[0054] Furthermore, feedforward multilayer neural networks employ the backpropagation learning algorithm.

[0055] Example 2:

[0056] The following is combined with Figure 2 This embodiment describes a fault identification system for radar level gauges used for volatile substances. The echo graph analysis module uses a specific method to determine whether there are crystalline deposits on the antenna based on the fluctuation values ​​on the echo graph. When one echo curve shows a fluctuation value while the other echo curves fluctuate smoothly, it is determined that there are crystalline deposits on the antenna.

[0057] Furthermore, the fluctuation values ​​appearing on the echo curve are in the range of 1 / 3 to 2 / 3 of the peak value.

[0058] Furthermore, a smooth oscillation of the echo curve is defined as an oscillation value within 1 / 4 of the peak value.

[0059] Furthermore, the actual liquid level is the fluctuation value on the echo curve that is above 2 / 3 of the peak value.

[0060] In this embodiment, a simulation is performed with a radar level gauge pulse transmission time interval of 1 second and a frequency of 2.5MHz. The fitting curves of the classification results (radar signal strength) output by the echo signal extraction module and the echo times of the radar level gauge are plotted to obtain echo graphs for different classification results. The echo graphs for different classification results are shown below. Figure 2As shown, when one echo curve exhibits fluctuations within 1 / 3 to 2 / 3 of its peak value, while the other echo curves fluctuate smoothly, it is determined that there are crystalline deposits on the antenna. The fluctuation value above 2 / 3 of the peak value represents the actual liquid level. Figure 2 In the diagram, 'a' indicates the presence of crystalline deposits on the antenna, and 'b' indicates the actual liquid level.

[0061] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A fault identification system for radar level gauges used for volatile substances, characterized in that, It includes: Echo signal extraction module, echo signal fitting module, and echo map analysis module; The echo signal extraction module uses binary wavelet transform to perform wavelet decomposition on the echo signal of the radar level gauge, principal component analysis for decorrelation, Fisher linear classification function for feature extraction, and BP neural network classifier for feature classification, outputting the classification result. The echo signal fitting module is used to plot the classification results output by the echo signal extraction module and the fitting curve of the radar level gauge echo time, and obtain the echo map of the echo curve of different classification results. The echo image analysis module is used to determine whether there are crystalline deposits on the antenna based on the fluctuation values ​​on the echo image. The specific method for wavelet decomposition of the echo signal of the radar level gauge using binary wavelet transform includes: mapping the echo signal of the radar level gauge to wavelet functions of different scales to obtain the small-scale information and large-scale information of the echo signal respectively. The specific methods for decorrelation using principal component analysis include: S3-1. Convert the small-scale information and large-scale information into input data vectors respectively; S3-2. Obtain the correlation matrix of the small-scale information input data vector, and then obtain the eigenvalues ​​and eigenvectors of the small-scale information; Obtain the correlation matrix of the large-scale information input data vector, and then obtain the eigenvalues ​​and eigenvectors of the large-scale information; S3-3. Select M dominant feature values, and the feature vectors corresponding to the M dominant feature values ​​constitute the feature subspace; where M is a positive integer; S3-4. Project the feature vectors of the small-scale information obtained in S3-2 onto the feature subspace obtained in S3-3; Project the feature vectors of the large-scale information obtained in S3-2 onto the feature subspace obtained in S3-3; Principal component analysis was used to decorrelate and compress high-dimensional data. The specific method for feature extraction using the Fisher linear classification function includes: Fisher linear classifier was used to sort the decorrelated and compressed data in descending order; Calculate the classification error of the Fisher linear classifier for different datasets; Based on the criterion of minimizing classification error, the values ​​with the greatest class discrimination ability are selected as the optimal feature subset and input into the neural network classifier.

2. The fault identification system for radar level gauges used for volatile substances according to claim 1, characterized in that, The specific method for classifying features using a BP neural network classifier includes: using a feedforward multilayer neural network to classify the input data.

3. The fault identification system for radar level gauges used for volatile substances according to claim 2, characterized in that, The feedforward multilayer neural network employs a backpropagation learning algorithm.

4. The fault identification system for radar level gauges used for volatile substances according to claim 1, characterized in that, The specific method for the echo graph analysis module to determine whether there are crystalline deposits on the antenna based on the fluctuation values ​​on the echo graph includes: when one echo curve shows a fluctuation value and the other echo curves fluctuate smoothly, it is determined that there are crystalline deposits on the antenna.

5. The fault identification system for radar level gauges used for volatile substances according to claim 4, characterized in that, The fluctuation values ​​appearing on the echo curve are in the range of 1 / 3 to 2 / 3 of the peak value.

6. The fault identification system for radar level gauges used for volatile substances according to claim 4, characterized in that, The smooth fluctuation of the echo curve is defined as the fluctuation value being within 1 / 4 of the peak value.

7. The fault identification system for radar level gauges used for volatile substances according to claim 4, 5, or 6, characterized in that, The actual liquid level is indicated by fluctuations in the echo curve that are above 2 / 3 of the peak value.

Images