Suspended sediment content measurement method based on KAF-RBF synergistic fusion
Through KAF-RBF collaborative fusion technology, the core adaptive filter and RBF neural network model are used to solve the noise and environmental interference problems in river sand content measurement, the measurement accuracy and applicable interval are improved, and the error in the existing technology is solved The problem of large and narrow application scope.
Patent Information
- Application Number
- CN202210457653.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-04-28
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2042-04-28
AI Technical Summary
The existing river sand content measurement methods have large errors, large labor consumption, long testing time and safety hazards. The capacitance voltage method is easily affected by temperature. The vibration method is not suitable for river sediment measurement, and the ultrasonic method is only suitable for river sediment measurement. If the amount of sand is small, the measurement error is large.
Using the suspension sand content measurement method based on KAF-RBF collaborative fusion, an ultrasonic measurement system is constructed, and a nuclear adaptive filter is used to eliminate noise interference, combined with inversion algorithm and RBF neural network model, environmental interference is eliminated and measurement accuracy is improved. and applicable range.
Effectively eliminate instrument noise and environmental interference, improve the river sediment measurement accuracy and applicable range of ultrasonic methods, reduce measurement errors, and expand the scope of application of sediment particle size.
Smart Images

Figure CN114965199B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of river sediment content measurement, and in particular to a suspended sediment content measurement method based on KAF-RBF collaborative fusion. Background Art
[0002] The sediment content of a river is one of the important bases for the design of hydraulic structures and water resources planning. Measuring the sediment in a river is also an important task for hydrological and water resources observation. At present, most hydrological observation stations at home and abroad use instantaneous samplers to measure the sediment in a river. However, such samplers generally have problems such as large errors, large labor consumption, long test time, and potential safety hazards. In recent years, technologies such as capacitance differential pressure method, vibration method, and ultrasonic wave have been widely used in the measurement of sediment content. The capacitance method is easily affected by temperature. In actual measurement, flow velocity and the like will also cause great interference to the measurement results, and the applicable conditions are very limited. The vibration method has a large measuring instrument volume and is affected by water depth and flow velocity, so it is not suitable for sediment measurement in a river. Ultrasonic technology performs relatively well in the measurement of suspended sediment in a river, but this method is only applicable to the case where the sediment content in a river is small. When the sediment content is large, the measurement error is large. In view of these problems, this application proposes a solution. Summary of the invention
[0003] Purpose of the invention: The purpose of the present invention is to provide a method for measuring suspended sediment content based on KAF-RBF collaborative fusion, which can eliminate the noise interference of the instrument itself and the interference of environmental factors on the measurement of sediment, effectively improve the applicable range of ultrasonic sediment content measurement and improve the measurement accuracy.
[0004] Technical solution: The suspended sediment content measurement method based on KAF-RBF collaborative fusion described in the present invention specifically includes the following steps:
[0005] S1: construct an ultrasonic measurement system, which includes a detection module, a noise removal module, a data fusion processing module and an output module;
[0006] S2: The river flow signal amplitude M(n) collected by the ultrasonic probe in the detection module is processed by a kernel adaptive filter to remove the noise interference of the sensor itself in the received signal amplitude M(n) and obtain the processed real signal amplitude M t (n);
[0007] S3: The real signal amplitude M t (n) Use the inversion algorithm to preliminarily obtain river sediment information: sediment content per unit volume S and sediment particle size r;
[0008] S4: Input the initially acquired river sediment information and the measurement area distance d, water depth h, water temperature T and flow velocity V into the RBF neural network model for training. The hidden layer to the output layer in the RBF neural network model is a linear mapping. The weight of the mapping function is solved by a linear function. The basis function of the RBF neural network model is a Gaussian function:
[0009]
[0010] where φ j is the jth node output in the hidden layer, C j is the center value of the Gaussian basis function, the number of hidden layer nodes is M, and σ is a normalization constant;
[0011] The output function of the linear combination of the node outputs of the hidden layer through the output layer is:
[0012]
[0013] Where p is the number of nodes, w ij Give the network output a weighted value;
[0014] S5: The RBF neural network model uses the error between the output value and the measured value, and takes the minimum mean square error as the principle to calculate the weight of the output layer, and finally obtains the sand content value S after eliminating noise interference and environmental interference. t .
[0015] Preferably, the detection module in S1 includes an ultrasonic probe, a water temperature meter, a flow meter, a water depth meter and a pressure meter, and the ultrasonic probe is a multi-frequency ultrasonic probe.
[0016] As a preference, the real signal amplitude M in S2 t The specific solution of (n) is as follows: the transmitting frequency of the ultrasonic probe is f, the signal amplitude of the input kernel adaptive filter is M(n), the expected signal (the expectation of the interference signal) is D(n), and the output value of D(n) after passing through the kernel adaptive filter is y(n). The error e(n) is obtained by subtracting M(n) from y(n) and then re-input into the kernel adaptive filter, and the weight coefficient and vector parameters are automatically updated. When the error converges, the output value y(n) approaches the real noise, so that e(n) approaches the real signal amplitude M. t (n).
[0017] Preferably, the frequency that the kernel adaptive filter in S2 needs to limit is the same as the transmission frequency of the ultrasonic probe.
[0018] Preferably, the method of preliminarily obtaining the river sediment information by using the inversion algorithm in S3 specifically includes the following steps:
[0019] S3.1: Since ultrasonic waves will attenuate when propagating in the medium, calculate the change in sound pressure:
[0020] I=I0e -ax
[0021] Where I0 is the initial sound pressure, a is the sound attenuation coefficient, and x is the propagation distance;
[0022] S3.2: Calculate the acoustic attenuation coefficient using the first echo A1 and the second echo A2 of the ultrasonic probe in the measurement area:
[0023] M1=e -2αpdp R p,s M i
[0024] M2=e -2αpdp (1-R 2 p,s )R s,w M i e -2αd
[0025] Where M1 represents the amplitude of the first echo signal, M2 represents the amplitude of the second echo signal, and M i represents the original signal amplitude, a represents the sound attenuation coefficient, d represents the distance of the measurement area, R p,s It represents the transmission coefficient from the ultrasonic probe glass to the area to be tested, R s,w It represents the reflection coefficient from the test area to the reflector, from which the sound attenuation coefficient a can be expressed as follows:
[0026]
[0027] S3.3: Change the frequency of the ultrasonic probe to calculate different sound attenuation coefficients a, obtain the ultrasonic attenuation spectrum, and thus obtain the distribution of the sediment particle size r;
[0028] S3.4: According to the sound pressure variation equation, obtain preliminary sediment content information:
[0029]
[0030] Where S is the sediment content per unit volume, M i represents the original signal amplitude, d represents the distance of the measurement area, represents the scattering correction factor, k s The particle scattering coefficient is determined by the particle size, k t represents the system parameters, and a represents the sound attenuation coefficient.
[0031] Beneficial effects: This application can eliminate noise interference on the instrument and reduce the impact of environmental factors on the measurement, thereby further improving the accuracy of ultrasonic prediction of river sediment and expanding the applicable range of sediment particle size. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] Figure 1 It is a flowchart of this application;
[0033] Figure 2 This is the architecture diagram of the Concord model in this application; DETAILED DESCRIPTION
[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0035] like Figure 1-2 As shown, this application specifically includes the following steps:
[0036] S1: construct an ultrasonic measurement system, which includes a detection module, a noise removal module, a data fusion processing module and an output module;
[0037] The detection module includes an ultrasonic probe, a water temperature meter, a flow meter, a water depth meter and a pressure gauge. In this embodiment, the ultrasonic probe adopts a multi-frequency ultrasonic probe;
[0038] S2: The river flow signal amplitude M(n) collected by the ultrasonic probe in the detection module is processed by a kernel adaptive filter to remove the noise interference of the sensor itself in the received signal amplitude M(n) and obtain the processed real signal amplitude M t (n);
[0039] The limiting frequency of the kernel adaptive filter is the same as the transmitting frequency of the ultrasonic probe, specifically:
[0040] The transmitting frequency of the ultrasonic probe is f, the signal amplitude of the input kernel adaptive filter is M(n), the expected signal (the expectation of the interference signal) is D(n), and the output value of D(n) after passing through the kernel adaptive filter is y(n). The error e(n) is obtained by subtracting M(n) from y(n) and then re-input into the kernel adaptive filter, and the weight coefficient and vector parameters are automatically updated. When the error converges, the output value y(n) approaches the real noise, so that e(n) approaches the real signal amplitude M. t (n)
[0041] S3: The real signal amplitude M t (n) Using the inversion algorithm to preliminarily obtain river sediment information: sediment content per unit volume S and sediment particle size r, specifically including the following steps:
[0042] S3.1: Since ultrasonic waves will attenuate when propagating in the medium, calculate the change in sound pressure:
[0043] I=I0e -ax
[0044] Where I0 is the initial sound pressure, a is the sound attenuation coefficient, and x is the propagation distance;
[0045] S3.2: Calculate the acoustic attenuation coefficient using the first echo A1 and the second echo A2 of the ultrasonic probe in the measurement area:
[0046] M1=e -2αpdp R p,s M i
[0047] M2=e -2αpdp (1-R 2 p,s )R s,w M i e -2αd
[0048] Where M1 represents the amplitude of the first echo signal, M2 represents the amplitude of the second echo signal, and M i represents the original signal amplitude, a represents the sound attenuation coefficient, d represents the distance of the measurement area, R p,s It represents the transmission coefficient from the ultrasonic probe glass to the area to be tested, R s,w It represents the reflection coefficient from the test area to the reflector, from which the sound attenuation coefficient a can be expressed as follows:
[0049]
[0050] S3.3: Change the frequency of the ultrasonic probe to calculate different sound attenuation coefficients a, obtain the ultrasonic attenuation spectrum, and thus obtain the distribution of the sediment particle size r;
[0051] S3.4: According to the sound pressure variation equation, obtain preliminary sediment content information:
[0052]
[0053] Where S is the sediment content per unit volume, M i represents the original signal amplitude, d represents the distance of the measurement area, represents the scattering correction factor, k s The particle scattering coefficient is determined by the particle size, k t represents the system parameters, a represents the sound attenuation coefficient;
[0054] S4: Input the initially acquired river sediment information and the measurement area distance d, water depth h, water temperature T and flow velocity V into the RBF neural network model for training. The hidden layer to the output layer in the RBF neural network model is a linear mapping. The weight of the mapping function is solved by a linear function. The basis function of the RBF neural network model is a Gaussian function:
[0055]
[0056] where φ j is the jth node output in the hidden layer, C j is the center value of the Gaussian basis function, the number of hidden layer nodes is M, and σ is a normalization constant;
[0057] The output function of the linear combination of the node outputs of the hidden layer through the output layer is:
[0058]
[0059] Where p is the number of nodes, w ij Give the network output a weighted value;
[0060] S5: The RBF neural network model uses the error between the output value and the measured value, and takes the minimum mean square error as the principle to calculate the weight of the output layer, and finally obtains the sand content value S after eliminating noise interference and environmental interference. t .
Claims
1. The suspended sediment content measurement method based on KAF-RBF synergistic fusion is characterized by: The specific steps include: S1: construct an ultrasonic measurement system, which includes a detection module, a noise removal module, a data fusion processing module and an output module; S2: The river flow signal amplitude M(n) collected by the ultrasonic probe in the detection module is processed by a kernel adaptive filter to remove the noise interference of the sensor itself in the received signal amplitude M(n) and obtain the processed real signal amplitude M t (n); S3: for the real signal amplitude M t (n) Use the inversion algorithm to preliminarily obtain river sediment information: sediment content per unit volume S and sediment particle size r. The specific steps are as follows: S3.1: Since ultrasonic waves will attenuate when propagating in the medium, calculate the change in sound pressure: I=I0e -ax Where I0 is the initial sound pressure, a is the sound attenuation coefficient, and x is the propagation distance; S3.2: Calculate the acoustic attenuation coefficient using the first echo A1 and the second echo A2 of the ultrasonic probe in the measurement area: M1=e -2αpdp R p,s M i M2=e -2αpdp (1-R 2 p,s )R s,w M i e -2αd Where M1 represents the amplitude of the first echo signal, M2 represents the amplitude of the second echo signal, and M i represents the original signal amplitude, a represents the sound attenuation coefficient, d represents the distance of the measurement area, R p,s It represents the transmission coefficient from the ultrasonic probe glass to the area to be tested, R s,w It represents the reflection coefficient from the test area to the reflector, from which the sound attenuation coefficient a can be expressed as follows: S3.3: Change the frequency of the ultrasonic probe to calculate different sound attenuation coefficients a, obtain the ultrasonic attenuation spectrum, and thus obtain the distribution of the sediment particle size r; S3.4: According to the sound pressure variation equation, obtain preliminary sediment content information: Where S is the sediment content per unit volume, M i represents the original signal amplitude, d represents the distance of the measurement area, represents the scattering correction factor, k s The particle scattering coefficient is determined by the particle size, k t represents the system parameters, a represents the sound attenuation coefficient; S4: Input the initially acquired river sediment information and the measurement area distance d, water depth h, water temperature T and flow velocity V into the RBF neural network model for training. The hidden layer to the output layer in the RBF neural network model is a linear mapping. The weight of the mapping function is solved by a linear function. The basis function of the RBF neural network model is a Gaussian function: where φ j is the jth node output in the hidden layer, C j is the center value of the Gaussian basis function, the number of hidden layer nodes is M, and σ is a normalization constant; The output function of the linear combination of the node outputs of the hidden layer through the output layer is: Where p is the number of nodes, w ij Give the network output a weighted value; S5: The RBF neural network model uses the error between the output value and the measured value, and takes the minimum mean square error as the principle to calculate the weight of the output layer, and finally obtains the sand content value S after eliminating noise interference and environmental interference. t .
2. The suspended sediment content measurement method based on KAF-RBF collaborative fusion according to claim 1 is characterized in that: The detection module in S1 includes an ultrasonic probe, a water temperature meter, a flow meter, a water depth meter and a pressure gauge, and the ultrasonic probe is a multi-frequency ultrasonic probe.
3. The suspended sediment content measurement method based on KAF-RBF collaborative fusion according to claim 1 is characterized in that: The real signal amplitude M in S2 t The specific solution of (n) is as follows: the transmitting frequency of the ultrasonic probe is f, the signal amplitude of the input kernel adaptive filter is M(n), the expected signal is the expected interference signal D(n), and the output value of D(n) after passing through the kernel adaptive filter is y(n). The error e(n) is obtained by subtracting M(n) from y(n) and then re-input into the kernel adaptive filter, and the weight coefficient and vector parameter are automatically updated. When the error converges, the output value y(n) approaches the real noise, so that e(n) approaches the real signal amplitude M. t (n).
4. The suspended sediment content measurement method based on KAF-RBF collaborative fusion according to claim 1 is characterized in that: The frequency that the kernel adaptive filter in S2 needs to be limited to is the same as the transmitting frequency of the ultrasonic probe.
